{ "schema_version": 1, "requested_window": { "start": "2026-10-08", "end": "2026-10-12" }, "review_finalized_on": "2026-10-12", "review_as_of": "2026-10-12T12:55:18+08:00", "candidate_review_complete": true, "all_sources_exhaustive": false, "implementation_complete": false, "unique_papers": 1610, "counts": { "accepted": 520, "rejected_or_out_of_scope": 424, "reviewed_deferred": 618, "known": 48 }, "coverage_counts": { "queried_unique": 1597, "inherited_in_query": 1019, "fresh_queried_reviewed": 578, "manual_supplement_outside_recall": 10, "inherited_outside_this_recall": 3 }, "limitations": [ "Only configured four-track two-mode retrieval and named official source scope; not all-web exhaustiveness.", "Bounded original defining methods and evaluation excerpts, not every appendix or full article.", "Supplementary discoveries are separately counted, never inflated into queried recall.", "Review closure does not mean all accepted historical candidates were implemented." ], "papers": [ { "id": "2610.13935", "title": "Hide and Seek: HIDE Concepts to SEEK explanations for black box models", "published": "2026-10-09T17:56:59Z", "reviews": [ { "id": "2610.13935", "title": "Hide and Seek: HIDE Concepts to SEEK explanations for black box models", "published": "2026-10-09T17:56:59Z", "decision": "reviewed_deferred", "reason": "给定concept文本，经VLM检索concept-aligned真实图像并localize/mask对应区域；black-box model原图/遮挡图输出差构成global/local concept attribution，非梯度weight-access方法。; CelebA/COCO-Stuff与三视觉classifier、captioning与crop segmentation应用；concept embedding preprocessing有不同VLM消融。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13935v1#S3", "https://arxiv.org/html/2610.13935v1#S4", "https://arxiv.org/html/2610.13935v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13935v1", "sha256": "ae0e220f190e59178cc088d065e17bf1f7d76848d388c05aa1e30bad4e78c68e" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13925", "title": "Can machines reason non-verbally?", "published": "2026-10-09T17:54:36Z", "reviews": [ { "id": "2610.13925", "title": "Can machines reason non-verbally?", "published": "2026-10-09T17:54:36Z", "decision": "reviewed_deferred", "reason": "GenVerOpt按sample生成多candidate、task-specific verifier评估并optimization/refinement，四task分别实例化，不能只复用抽象Generate-Verify口号。; visual IQ和三medical image case studies与human/模型比较；本轮只审bounded recipe/主实验段，未完整核四个实例/validation。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13925v1#Sx1", "https://arxiv.org/html/2610.13925v1#Sx2" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13925v1", "sha256": "4dcb4a6c0955369006da7e0d7c5166515b9db410a44ab916834d65d1e294f51f" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13917", "title": "Designing Protein Structures from Natural Language", "published": "2026-10-09T17:51:55Z", "reviews": [ { "id": "2610.13917", "title": "Designing Protein Structures from Natural Language", "published": "2026-10-09T17:51:55Z", "decision": "reviewed_deferred", "reason": "Prompteina用1.8M structure/50M+captions对pretrained all-atom Proteina-Complexa text conditioning训练，residue-text对齐finetune；SDE+CFG生成backbone/sidechainlatents，frozenVAE解码结构/序列。; structural properties、20binder targets和iterative in-silico prompt revision；GM2A实际89designs SPR assay发现9binders。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13917v1#S2", "https://arxiv.org/html/2610.13917v1#S3", "https://arxiv.org/html/2610.13917v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13917v1", "sha256": "d6d74fb9decde0583f0addccedc7c60e2fdb131a9e82a72ac1e2a11f762afbf0" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13912", "title": "RSI-Anything: Unlocking Recursive Self-Improvement with World Models", "published": "2026-10-09T17:50:30Z", "reviews": [ { "id": "2610.13912", "title": "RSI-Anything: Unlocking Recursive Self-Improvement with World Models", "published": "2026-10-09T17:50:30Z", "decision": "accepted", "reason": "用户点名冻结16范围；selected defining-method/experiment原文证据核验，具体实现/运行等级独立记录。; world model预测候选行动后外部执行验证；真实policy/workflow更新与按失配触发predictor适配，多轮component gate后整对compatibility retention。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "实施期间读取摘要、中心方法/训练/评测及部分相关附录；本次复用这些已读来源并重新校验官方HTML字节SHA和v1日期，不声明全文每段已读。", "review_artifact": "research-audits/oct12-review-closure/frozen-agent-reviews.json", "evidence_locations": [ "2.1 Problem formulation", "2.2 Predictive foundation", "2.3 Prediction-guided selection", "2.4 Feedback-driven improvement", "2.5 Recursive retention", "3.1 Code repair", "3.2 GUI environments", "3.3 Laboratory planning", "3.4 Reaction optimization", "3.5 Immune-science application", "6.1 Domain-specific evaluation settings", "6.2 Statistical reporting" ], "dateproof": { "url": "https://arxiv.org/abs/2610.13912", "sha256": "96f3e90d064e3616a965e56cec01c28263d933e2306dcc11ab57bf4b1baf7e5c", "location": "Submission history / [v1]", "first_publication_utc": "2026-10-09T17:50:30Z" }, "source_proof": { "url": "https://arxiv.org/html/2610.13912v1", "sha256": "af2f040a4871749019a14acad424bcb5d0c14865f95ed0e283b495261856c1ec", "byte_hash_verified_on": "2026-10-12" }, "new_implementation_inferred_from_review": false } ], "discovery_origin": "independent-targeted-supplement", "implementation_status": "registered_scoped_runtime; actual acceptance tracked separately", "status": "accepted", "implementation_key": "rsi-anything", "accepted_scope": "bounded faithful mechanism/runtime or evaluation infrastructure; not original capability replication" }, { "id": "2610.13910", "title": "When VLMs Destroy What They Know: Spatial Representations Fall Below Lexical Priors", "published": "2026-10-09T17:50:03Z", "reviews": [ { "id": "2610.13910", "title": "When VLMs Destroy What They Know: Spatial Representations Fall Below Lexical Priors", "published": "2026-10-09T17:50:03Z", "decision": "reviewed_deferred", "reason": "CASP-VL在VisualGenome scenegraph只改一个concept/rendercaption；linearprobe仅原caption训练，EICS问perturbedcaption是否随改动稳定解码，BoW作为lexical baseline。; 跨CLIP/SigLIP/LLaVA layers、predicate-token与EOS/mean pooling、outputbenchmark对照及bootstrapCI。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13910v1#S3", "https://arxiv.org/html/2610.13910v1#S4", "https://arxiv.org/html/2610.13910v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13910v1", "sha256": "9dd1425bbab2c1f11831e4eb49d40fb3f64fd822a6a3381d260221f72a2bfbf7" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13896", "title": "Structural Automata for Neural Semantic Parsing", "published": "2026-10-09T17:42:14Z", "reviews": [ { "id": "2610.13896", "title": "Structural Automata for Neural Semantic Parsing", "published": "2026-10-09T17:42:14Z", "decision": "reviewed_deferred", "reason": "Structural automaton rejected-token training与APD保证固定length budget下accepting path；PMB train/dev/test清楚并dev loss选checkpoint，下一批需真实automaton completion witness与DRS/AMR parser，不用gold完成fixture。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 One Automaton, Three Stages", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experimental Setup", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13896v1", "sha256": "4cdb01c4e378eae3305eca830fe266aa17a2e5cd6dd0c930519d0d775e4e5a5d", "local_html_sha256": "4cdb01c4e378eae3305eca830fe266aa17a2e5cd6dd0c930519d0d775e4e5a5d", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13896v1", "result": "retrieved", "sha256": "4cdb01c4e378eae3305eca830fe266aa17a2e5cd6dd0c930519d0d775e4e5a5d" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.13889", "title": "RATED: Reliability-Aware AI Agent Tool Retrieval via Execution Dynamics", "published": "2026-10-09T17:38:11Z", "reviews": [ { "id": "2610.13889", "title": "RATED: Reliability-Aware AI Agent Tool Retrieval via Execution Dynamics", "published": "2026-10-09T17:38:11Z", "decision": "accepted", "reason": "用户点名冻结16范围；selected defining-method/experiment原文证据核验，具体实现/运行等级独立记录。; 独立语义双encoder真实InfoNCE训练后冻结，semantic shortlist融合prior与discounted contextual LinUCB，只按实际被选tool的执行/失败更新arm。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "实施期间读取摘要、中心方法/训练/评测及部分相关附录；本次复用这些已读来源并重新校验官方HTML字节SHA和v1日期，不声明全文每段已读。", "review_artifact": "research-audits/oct12-review-closure/frozen-agent-reviews.json", "evidence_locations": [ "3.2.1 DRAGON Dual-Encoder Architecture", "3.2.2 Contrastive Fine-tuning on EcoAgent-Syn", "3.3 Reliability-Aware Re-ranking", "3.4 Hybrid Fusion for Tool Ranking", "2 The EcoAgent Benchmark", "4.1 Experimental Setup", "4.2 Evaluation Metrics", "4.4 Main Results", "4.5 Hyperparameter Sensitivity", "4.6 Efficiency Analysis" ], "dateproof": { "url": "https://arxiv.org/abs/2610.13889", "sha256": "63a0da2b6117f86761fc9ddefd21081e5a3ad667434773a405915d655c895bd4", "location": "Submission history / [v1]", "first_publication_utc": "2026-10-09T17:38:11Z" }, "source_proof": { "url": "https://arxiv.org/html/2610.13889v1", "sha256": "33f3eea26c06ae5f417ea95098d0533bf82b7bfdf176c9af94b3c0d270684afb", "byte_hash_verified_on": "2026-10-12" }, "new_implementation_inferred_from_review": false } ], "discovery_origin": "independent-targeted-supplement", "implementation_status": "registered_scoped_runtime; actual acceptance tracked separately", "status": "accepted", "implementation_key": "rated", "accepted_scope": "bounded faithful mechanism/runtime or evaluation infrastructure; not original capability replication" }, { "id": "2610.13886", "title": "X-AgentRed: Red-Teaming Personal Agents in Evolving Multi-Service Ecosystems", "published": "2026-10-09T17:34:44Z", "reviews": [ { "id": "2610.13886", "title": "X-AgentRed: Red-Teaming Personal Agents in Evolving Multi-Service Ecosystems", "published": "2026-10-09T17:34:44Z", "decision": "reviewed_deferred", "reason": "Meta 优先。HarmFlow将工具前提/效果/危害建成有向图，反向采样并以真实执行反馈替换工具和修复阻塞前提；已读§3与§4缓存段。ToolSimulator30服务914工具、4模型的有害路径/覆盖率评测，不是生产A/B。下一批须隔离工具权限、复核双judge，依赖链接不能冒认为本论文开源。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 X-AgentRed", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 11280 ] }, { "heading": "4 Experiments", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 8456 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13886v1", "sha256": "8d1aabd630e8984cdaf921231fe8db529191987c74811a3fb8d2cbd0bf7e15ca", "local_html_sha256": "8d1aabd630e8984cdaf921231fe8db529191987c74811a3fb8d2cbd0bf7e15ca", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13886v1", "result": "retrieved", "sha256": "8d1aabd630e8984cdaf921231fe8db529191987c74811a3fb8d2cbd0bf7e15ca" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13858", "title": "LapMerge: Differentiable Exact-Budget Token Merging in Linear Time", "published": "2026-10-09T17:07:17Z", "reviews": [ { "id": "2610.13858", "title": "LapMerge: Differentiable Exact-Budget Token Merging in Linear Time", "published": "2026-10-09T17:07:17Z", "decision": "reviewed_deferred", "reason": "Laplace mixture CDF把ordered tokens按equal mass分成exact N intervals；prefix/suffix闭式barrier+feature aggregation和backward，不构造M×N；log-space避免指数overflow。; §4 GH200大序列runtime、patient-level6seedTCGA、Qwen2.5VL POPE frozen replacement、3seedLRA。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract/author information/main defining methods/main experiments and tables; production-evidence paragraph spot check; appendices not exhaustively read", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13858v1#S3", "https://arxiv.org/html/2610.13858v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13858v1", "sha256": "0ef73f65845cdbc22fd590c855c5b166de110953ac63ff903518c38cc8d6b778" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13841", "title": "Returning to Development States: When Full Sandbox Snapshots Help Coding Agents", "published": "2026-10-09T16:50:19Z", "reviews": [ { "id": "2610.13841", "title": "Returning to Development States: When Full Sandbox Snapshots Help Coding Agents", "published": "2026-10-09T16:50:19Z", "decision": "reviewed_deferred", "reason": "已读摘要：Daytona完整机器快照的Go-Explore搜索，收益依赖仓库外数据库/服务状态，区别于代码diff重试。官方正文缓存不可用，方法与控制组尚未核验，保留未决而非关键词拒绝。", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "read_scope": "abstract_only_primary_fulltext_unavailable", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "url": "https://arxiv.org/abs/2610.13841", "sha256": null, "local_html_sha256": null, "local_html_sha256_matches": false, "attempts": [ { "url": "https://arxiv.org/html/2610.13841v1", "result": "failed", "error": "404 Client Error: Not Found for url: https://arxiv.org/html/2610.13841v1" }, { "url": "https://arxiv.org/html/2610.13841v1", "result": "failed", "error": "404 Client Error: Not Found for url: https://arxiv.org/html/2610.13841v1" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13840", "title": "Studying Comics at Scale with Artificial Intelligence: The PAGE Framework and Three Studies of American Comics and Japanese Manga", "published": "2026-10-09T16:49:29Z", "reviews": [ { "id": "2610.13840", "title": "Studying Comics at Scale with Artificial Intelligence: The PAGE Framework and Three Studies of American Comics and Japanese Manga", "published": "2026-10-09T16:49:29Z", "decision": "rejected_or_out_of_scope", "reason": "PAGE作为comics研究instrument：parse panel/OCR/bibliography、固定codebook VLM annotation、独立human/reference grounding与genre/publisher controlled趋势估计。; 3929Americanbooks/87manga；name-withheld人标gender核验，修正图文页码偏移后retrieval比dense增益CI跨零。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13840v1", "sha256": "9f20dfdb7de9af74a03125fa07cef537f64700b8932c947f5debb84ac4535e9a" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.13832", "title": "SpatialASM: Internalizing Active Mentalization for Spatial Reasoning", "published": "2026-10-09T16:39:50Z", "reviews": [ { "id": "2610.13832", "title": "SpatialASM: Internalizing Active Mentalization for Spatial Reasoning", "published": "2026-10-09T16:39:50Z", "decision": "reviewed_deferred", "reason": "训练时真实allocentric novel-view generator空间模拟；LVLM view-plan/NVG render后转caption的ASM轨迹SFT，自动perspective-taking tasks和GeoP几何一致性约束。; 四LVLM backbone、7spatial+3generalVQA；外置ASM与内部SFT/RL消融，单SFT/RL优势随task不同。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13832v1#S3", "https://arxiv.org/html/2610.13832v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13832v1", "sha256": "d95d3fc51bc40f356391bd959e8a39aca4082e0dbca7af461669ad357d3fbcb6" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13821", "title": "Can Language Models Learn When to Abstain? Policy Learning and Calibrated Answer Selection", "published": "2026-10-09T16:33:27Z", "reviews": [ { "id": "2610.13821", "title": "Can Language Models Learn When to Abstain? Policy Learning and Calibrated Answer Selection", "published": "2026-10-09T16:33:27Z", "decision": "reviewed_deferred", "reason": "Qwen3-14B factual abstention比较RL policy和trained selector，独立training/calibration与heldout thresholds；selector TriviaQA改善但NQ仍不及always abstain，下一批保留negative transfer和utility baseline。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Results", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13821v1", "sha256": "714264f7afe089a6666ad89798fb294a4f5e522467cf6a434fd2b965fc45e3bd", "local_html_sha256": "714264f7afe089a6666ad89798fb294a4f5e522467cf6a434fd2b965fc45e3bd", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13821v1", "result": "retrieved", "sha256": "714264f7afe089a6666ad89798fb294a4f5e522467cf6a434fd2b965fc45e3bd" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.13806", "title": "FORTE: A Production Framework for Feature Evaluation and Deprecation in Industrial Recommender Systems", "published": "2026-10-09T16:18:39Z", "reviews": [ { "id": "2610.13806", "title": "FORTE: A Production Framework for Feature Evaluation and Deprecation in Industrial Recommender Systems", "published": "2026-10-09T16:18:39Z", "decision": "accepted", "reason": "Pinterest 原文§4各阶段与§7.1/7.2生产A/B逐项核对；48-feature上线、59-feature护栏失败；已实现的公开离线核心机制与生产结果分开。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "bounded_methods_and_evaluation", "review_artifact": "research-audits/oct12-review-closure/recommendation-reviews.json", "evidence_locations": [ "§3 Eq1", "§4.1–4.5", "§5.1", "§6.2", "§7.1 Table1", "§7.2 Table2" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13806v1", "sha256": "4e30ba8df33f09a72fa542cbfe96a67ba1d900336c150b228eada2ecb6dcce89" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "recommendation" ], "status": "accepted", "implementation_key": "forte", "accepted_scope": "bounded faithful mechanism/runtime or evaluation infrastructure; not original capability replication" }, { "id": "2610.13797", "title": "REFLECT: Role-Aware Detection of Sentiment, Emotion, Politeness, and Formality", "published": "2026-10-09T16:10:25Z", "reviews": [ { "id": "2610.13797", "title": "REFLECT: Role-Aware Detection of Sentiment, Emotion, Politeness, and Formality", "published": "2026-10-09T16:10:25Z", "decision": "reviewed_deferred", "reason": "BERT+theory-grounded SRL映射：CLS对role vectors的internal-state attention和ARG0→target directional external-cue attention，用learned gate融合context/role representation。; 四SEPF task officialtrain/dev/test，dev checkpoint选择与gridsearch；sentiment增益marginal非显著，emotion/politeness/formality较稳。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13797v1#Sx3", "https://arxiv.org/html/2610.13797v1#Sx4", "https://arxiv.org/html/2610.13797v1#Sx5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13797v1", "sha256": "cca38ceb36e59a86b83d93d11fece7fb136309daf76aa953f69781bb8e528e8c" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13790", "title": "OccludeBench: Can VLMs Find What's Missing?", "published": "2026-10-09T16:04:54Z", "reviews": [ { "id": "2610.13790", "title": "OccludeBench: Can VLMs Find What's Missing?", "published": "2026-10-09T16:04:54Z", "decision": "reviewed_deferred", "reason": "人工完全occlude对象，direct physical-cue与inferentialcontext两类；OccludeTrace四阶段诊断；实际image-generation inpaint candidate再与原图cue confront的干预。; 五VLM与human free-form，direct/inferential差距；reason/tool/fewshot无稳定消除，只有两imagegen模型尝试inpaint。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13790v1#S2", "https://arxiv.org/html/2610.13790v1#S3" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13790v1", "sha256": "d2c24230c46efe4acc58b06d61d9c18e6b026e38551b847ef03c151c7c1c0671" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13783", "title": "UkrMeme: Multimodal Explainable Persuasion Detection in Ukrainian Memes", "published": "2026-10-09T16:03:27Z", "reviews": [ { "id": "2610.13783", "title": "UkrMeme: Multimodal Explainable Persuasion Detection in Ukrainian Memes", "published": "2026-10-09T16:03:27Z", "decision": "rejected_or_out_of_scope", "reason": "200乌克兰wartime memes新标注及Propaganda950翻译inpaint/OCR文本替换数据；zero-shot三modality hierarchicalF1，LRP对照modeltokenattribution。; 四模型和训练label频率weightedrandom baseline，explanationtoken category由GPT4o分类并小样本人核。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13783v1#S3", "https://arxiv.org/html/2610.13783v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13783v1", "sha256": "c778a577218948beaa248eaff04ade8ffff0f88343d90e97fc50e146e1bfd958" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.13772", "title": "MARGoS: A Configuration-Driven Platform for Reproducible MARL Experiments in ARGoS", "published": "2026-10-09T15:55:16Z", "reviews": [ { "id": "2610.13772", "title": "MARGoS: A Configuration-Driven Platform for Reproducible MARL Experiments in ARGoS", "published": "2026-10-09T15:55:16Z", "decision": "rejected_or_out_of_scope", "reason": "已读摘要及平台/评测段：配置哈希、seed传递与ARGoS机器人MARL实验包，是传统机器人可复现基础设施；不是本批LLM agent算法或其公开执行评测，保留作工程参考。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "1 Project Context, Goals, and Scope", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 Platform Overview", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13772v1", "sha256": "82f3ace1ae47843e5c28fab7f776a259b1fa1ec529d94b0193084487b5ed48b0", "local_html_sha256": "82f3ace1ae47843e5c28fab7f776a259b1fa1ec529d94b0193084487b5ed48b0", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13772v1", "result": "retrieved", "sha256": "82f3ace1ae47843e5c28fab7f776a259b1fa1ec529d94b0193084487b5ed48b0" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.13767", "title": "DualQuant: Dual-scale Calibration-free Weight-and-Activation Quantization for LLMs", "published": "2026-10-09T15:51:38Z", "reviews": [ { "id": "2610.13767", "title": "DualQuant: Dual-scale Calibration-free Weight-and-Activation Quantization for LLMs", "published": "2026-10-09T15:51:38Z", "decision": "accepted", "reason": "black-boxquantizer、inner/outvariable splitting、闭式双inverse-scale投影坐标下降，weight-derivedactivationrescale再实际量化。; 已实现限定INT4/MXINT4software simulation，CPU公式单测通过；公开checkpoint GPU未完成。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract/author information/main defining methods/main experiments and tables; production-evidence paragraph spot check; appendices not exhaustively read", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13767v1#S4", "https://arxiv.org/html/2610.13767v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13767v1", "sha256": "eb28173622f2e939a056f0fb7aa00e472f9682cf13bfad6c3e49b56a690ac8b4" }, "new_implementation_inferred_from_review": false }, { "id": "2610.13767", "title": "DualQuant: Dual-scale Calibration-free Weight-and-Activation Quantization for LLMs", "published": "2026-10-09T15:51:38Z", "decision": "accepted", "reason": "weightreconstructionMSE非微分quantizerblackbox，variable splitting固定Q在innerloop闭式row/columninverse-scale coordinateupdates并格式可行集投影，outerrequantize。dual-scaleactivationfactorization从weightscale导出per-channelactivationrescaling而不用activationcalibration；推理同时W+A量化。; Llama/Qwen在WikiText2/C4/zero-shot、INT/MX/NV多格式与ablation；原文RTX4090packedW4A4path报告accuracy/memory/throughput，与本地referencepath分开。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "实施阶段已读摘要、作者信息、定义算法方法与主实验，以及列明的相关附录；复用实际实施审查，未宣称全部附录/整篇全文已读。", "review_artifact": "research-audits/oct12-review-closure/frozen-foundation-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13767v1", "sha256": "eb28173622f2e939a056f0fb7aa00e472f9682cf13bfad6c3e49b56a690ac8b4" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "accepted", "implementation_key": "dualquant", "accepted_scope": "bounded faithful mechanism/runtime or evaluation infrastructure; not original capability replication" }, { "id": "2610.13763", "title": "Neurons++: A Unified Visual-Stream Framework for Image and Video Reconstruction from fMRI with Neural-to-Visual Interpretability", "published": "2026-10-09T15:49:07Z", "reviews": [ { "id": "2610.13763", "title": "Neurons++: A Unified Visual-Stream Framework for Image and Video Reconstruction from fMRI with Neural-to-Visual Interpretability", "published": "2026-10-09T15:49:07Z", "decision": "reviewed_deferred", "reason": "image-fMRI multisubject pretrain→subject-specific image/video adaptation；ridge+residualMLP/CLIP+diffusionprior alignment，四stage serial temporaltransformer持续接原neuralrepresentation，依次mask/concept/scene/blurryview监督；unCLIP keyframe后modalityappropriateSDXL或T2V diffusion。; CC2017/HCP三subjects视频、NSD subjects1/2/5/7 images；HCP为作者group-aware temporal自建split，非官方test；ROI/randomgroupcontrol与delta-fMRI qualitative direction probes。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "official PDF pages1..7 first5500 extracted characters each; selected interpretability/limitations passages pages8..11; not complete25pages", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/pdf/2610.13763v1#page=3", "https://arxiv.org/pdf/2610.13763v1#page=5", "https://arxiv.org/pdf/2610.13763v1#page=7", "https://arxiv.org/pdf/2610.13763v1#page=8" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/pdf/2610.13763v1", "sha256": "1218f23bc0b65fc3c20d749075a71d5bb1d08f87a32e45bd77073542e3ef69ec" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13755", "title": "Matched-Context Group Policy Optimization for LLM Agent Training", "published": "2026-10-09T15:41:37Z", "reviews": [ { "id": "2610.13755", "title": "Matched-Context Group Policy Optimization for LLM Agent Training", "published": "2026-10-09T15:41:37Z", "decision": "reviewed_deferred", "reason": "MCGPO 用固定 sibling folds、state/rubric 匹配和反折 terminal evidence 构造 pairwise preference，3-seed ALFWorld/WebShop 控制支持后训练机制；冻结16篇之外，下一批需同状态/进度和label隔离实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13755v1", "sha256": "d68cc32d561fc07bb10a04b38a1fa7c443b1a50663f5dd3fdc8d200cba91d188", "local_html_sha256": "d68cc32d561fc07bb10a04b38a1fa7c443b1a50663f5dd3fdc8d200cba91d188", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13755v1", "result": "retrieved", "sha256": "d68cc32d561fc07bb10a04b38a1fa7c443b1a50663f5dd3fdc8d200cba91d188" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.13744", "title": "A Privileged Teacher That Only Scores Cannot Search: Critique-Conditioned On-Policy Self-Distillation for Reasoning", "published": "2026-10-09T15:36:47Z", "reviews": [ { "id": "2610.13744", "title": "A Privileged Teacher That Only Scores Cannot Search: Critique-Conditioned On-Policy Self-Distillation for Reasoning", "published": "2026-10-09T15:36:47Z", "decision": "reviewed_deferred", "reason": "Critique-conditioned OPSD先显式生成纠错critique再score，比privileged score-only错误定位更好；gold error spans只能外部诊断，test critique不可读取reference解；下一批需区分privileged training与gold-free test/runtime。 论文oracle critiques含reference只算privileged诊断；本地不能作为gold-free test能力证据。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "7 Critique-Conditioned Self-Distillation Delivers", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "4 Setup", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13744v1", "sha256": "bb77f821556417d32712123fba118abc815c7dd4105e0c8a4a5a3e668160db8e", "local_html_sha256": "bb77f821556417d32712123fba118abc815c7dd4105e0c8a4a5a3e668160db8e", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13744v1", "result": "retrieved", "sha256": "bb77f821556417d32712123fba118abc815c7dd4105e0c8a4a5a3e668160db8e" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.13737", "title": "SafeDiSTL: Decomposing Joint Constraints for Safe Multi-Agent Reinforcement Learning", "published": "2026-10-09T15:32:06Z", "reviews": [ { "id": "2610.13737", "title": "SafeDiSTL: Decomposing Joint Constraints for Safe Multi-Agent Reinforcement Learning", "published": "2026-10-09T15:32:06Z", "decision": "rejected_or_out_of_scope", "reason": "已读摘要及受约束Markov博弈/实验段：分布式时序逻辑安全约束与MAPPO/MACPO机器人MARL，不是LLM agent，本次冻结实现范围外。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Preliminaries", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Problem Formulation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13737v1", "sha256": "f9e9bc56d38e4ac0f69c8dbc53267cea93a5ad553e6773714b5f50d98bb8de31", "local_html_sha256": "f9e9bc56d38e4ac0f69c8dbc53267cea93a5ad553e6773714b5f50d98bb8de31", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13737v1", "result": "retrieved", "sha256": "f9e9bc56d38e4ac0f69c8dbc53267cea93a5ad553e6773714b5f50d98bb8de31" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.13735", "title": "Orthros: Phase-Aware Heterogeneous Attention for Efficient Transformers", "published": "2026-10-09T15:30:08Z", "reviews": [ { "id": "2610.13735", "title": "Orthros: Phase-Aware Heterogeneous Attention for Efficient Transformers", "published": "2026-10-09T15:30:08Z", "decision": "accepted", "reason": "fixed-windowbaseSWAprefill、outputtokenfullcausalcrossattention、同层完整KV与真实conversion/training。; 已实现并有CPU formula/cache训练smoke，GPU未完成；原文H100/longcontext结果与本地L1分开。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract/author information/main defining methods/main experiments and tables; production-evidence paragraph spot check; appendices not exhaustively read", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13735v1#S3", "https://arxiv.org/html/2610.13735v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13735v1", "sha256": "b7e186007a39f796527eb4a0ae02e1cabf842b6aa55f44f3a65de4cc439a41ae" }, "new_implementation_inferred_from_review": false }, { "id": "2610.13735", "title": "Orthros: Phase-Aware Heterogeneous Attention for Efficient Transformers", "published": "2026-10-09T15:30:08Z", "decision": "accepted", "reason": "phaseaware两stream共享attention参数：base固定窗口SWA产生线性prefill表示，output对base完整同层KV作fullcausalattention。prefill仅需最后output，decode真实state/cache保持完整baseKV；训练包含从scratch或Transformercheckpointconversion，不能kernelizedlinearattention代理SWA/fullcache。; 原文scaling、scratch、dat-efficientconversion、128kNIAH/realcontextrecall及prefill/end-to-end测时；21.9xprefill是source结果，不能套在本地tiny短context。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "实施阶段已读摘要、作者信息、定义算法方法与主实验，以及列明的相关附录；复用实际实施审查，未宣称全部附录/整篇全文已读。", "review_artifact": "research-audits/oct12-review-closure/frozen-foundation-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13735v1", "sha256": "b7e186007a39f796527eb4a0ae02e1cabf842b6aa55f44f3a65de4cc439a41ae" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "accepted", "implementation_key": "orthros", "accepted_scope": "bounded faithful mechanism/runtime or evaluation infrastructure; not original capability replication" }, { "id": "2610.13708", "title": "Adaptive DHiSparse: Speculative Decoding over a Load-Adaptive KV Hierarchy", "published": "2026-10-09T15:02:35Z", "reviews": [ { "id": "2610.13708", "title": "Adaptive DHiSparse: Speculative Decoding over a Load-Adaptive KV Hierarchy", "published": "2026-10-09T15:02:35Z", "decision": "reviewed_deferred", "reason": "requiredhotcache/tail与reclaimable optionalhistory分开；ready publication+host backup invariant；exact多querymiss dedup、不合并attention selections，draftwindowring与layergroupverificationstaging复用，acceptedprefix只commit已接受KV。; §4 8×B300，GLM5.3 100K与DeepSeekV4.1Flash500K concurrency serving，同model kernels，loadstep/width/hotcache sweeps与真实payloadcopy profiles。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract/author information/main defining methods/main experiments and tables; production-evidence paragraph spot check; appendices not exhaustively read", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13708v1#S3", "https://arxiv.org/html/2610.13708v1#S4", "https://arxiv.org/html/2610.13708v1#S4.SS1" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13708v1", "sha256": "ca614cd43e9e1c8ceb8ec50c98be73f552a35460a8011178cf61c66ebd28d1fc" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13703", "title": "A Baseline Assessment for the Agentic Web Technology Stack", "published": "2026-10-09T14:59:24Z", "reviews": [ { "id": "2610.13703", "title": "A Baseline Assessment for the Agentic Web Technology Stack", "published": "2026-10-09T14:59:24Z", "decision": "reviewed_deferred", "reason": "已读摘要与实验设置：MASTIF比较LLM、协议、框架组成的agentic web栈，300 Mind2Web任务，属于执行评测基础设施；需继续核验完整任务协议、成功评分及统计，下一批而非新算法实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3. Problem Definition", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4. Experiment Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13703v1", "sha256": "363952e103262fa17c485894e53b5997de0dff1c9d18b29e979d2f5902ca8c87", "local_html_sha256": "363952e103262fa17c485894e53b5997de0dff1c9d18b29e979d2f5902ca8c87", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13703v1", "result": "retrieved", "sha256": "363952e103262fa17c485894e53b5997de0dff1c9d18b29e979d2f5902ca8c87" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13697", "title": "Improving Semantic Coverage in AI Test & Evaluation through Generative Augmentation", "published": "2026-10-09T14:57:12Z", "reviews": [ { "id": "2610.13697", "title": "Improving Semantic Coverage in AI Test & Evaluation through Generative Augmentation", "published": "2026-10-09T14:57:12Z", "decision": "reviewed_deferred", "reason": "zero-shot VLM二元semantic因子+embedding clustering辨识coverage gap，diffusion image2image迁移语义保content，再按embedding drift/radius/completeness验证合格batch加入test；不重训被评模型。; CCT trailcam与BDD100K day/night toy gap，original/augmented/ideal真night测试对trained-day与diverse-trained模型区分。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13697v1#S3", "https://arxiv.org/html/2610.13697v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13697v1", "sha256": "4a717c5ca6768283ba69f3f06badae833b3990a7b1a14d3fff065d4f7b5c7b6d" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13690", "title": "Transformer-based In-Context Score Operator Learning for Diffusion Models", "published": "2026-10-09T14:53:49Z", "reviews": [ { "id": "2610.13690", "title": "Transformer-based In-Context Score Operator Learning for Diffusion Models", "published": "2026-10-09T14:53:49Z", "decision": "reviewed_deferred", "reason": "context empirical samples+noisy query/time输入query-dependent transformer score operator；Gaussian conditional-mean/Nadaraya-Watson exact-representable构造，跨task amortized denoising scorematch，unseen任务reverse diffusion不更新参数。; Navier-Stokes ID/OOD、ERA5/QuickDraw/Food101更多ID；复杂task退化为论文明确negative，theorem对指定time-modulated mean class而非任意Transformer。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13690v1#S3", "https://arxiv.org/html/2610.13690v1#S4", "https://arxiv.org/html/2610.13690v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13690v1", "sha256": "4f3625ff9247d3cdaff7b6bd07df28ee77c366dddbf2e9fc75a7e1f8338ee317" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13668", "title": "(When) Should a Software Factory be Lights-off? Benchmarking Code Review Agents on Bug Detection", "published": "2026-10-09T14:37:15Z", "reviews": [ { "id": "2610.13668", "title": "(When) Should a Software Factory be Lights-off? Benchmarking Code Review Agents on Bug Detection", "published": "2026-10-09T14:37:15Z", "decision": "reviewed_deferred", "reason": "已读摘要及§4缓存方法/实验：CRBugBench含公开与私有真实bug，固定Copilot CLI和推理设置，离线bug检出率与cluster-robust回归不等于生产A/B/因果作用；私有子集限制本地复现，下一批评测infra。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4. Experiments", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 16000 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13668v1", "sha256": "1c228b087c98034a9869ade681ca44750e844553543d883149c61d5d384773ef", "local_html_sha256": "1c228b087c98034a9869ade681ca44750e844553543d883149c61d5d384773ef", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13668v1", "result": "retrieved", "sha256": "1c228b087c98034a9869ade681ca44750e844553543d883149c61d5d384773ef" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13663", "title": "A Multi-Agent LLM System for Automated Tuning of Specialized Neural Network Models", "published": "2026-10-09T14:34:22Z", "reviews": [ { "id": "2610.13663", "title": "A Multi-Agent LLM System for Automated Tuning of Specialized Neural Network Models", "published": "2026-10-09T14:34:22Z", "decision": "reviewed_deferred", "reason": "固定网络/训练流水线，只让 Analyst提出指定模块规格、Coder生成契约约束代码、Docker实测、Recorder入log；错误保留但不计成功预算，不能把整个模型搜索当同一机制。; SNN MNIST/Digits与三PINN任务；正文区分best-run、物理守恒诊断与十次初始化重复。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13663v1#S3", "https://arxiv.org/html/2610.13663v1#S4", "https://arxiv.org/html/2610.13663v1#S5", "https://arxiv.org/html/2610.13663v1#S8" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13663v1", "sha256": "03f1c51af165fb192abb1f13eff3038977f22e81c4a9a8836183329e0d84eb20" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13660", "title": "Pretraining and Test-Time Training Budgets Matter for ARC: A Controlled Study of VARC and TRM", "published": "2026-10-09T14:33:30Z", "reviews": [ { "id": "2610.13660", "title": "Pretraining and Test-Time Training Budgets Matter for ARC: A Controlled Study of VARC and TRM", "published": "2026-10-09T14:33:30Z", "decision": "reviewed_deferred", "reason": "在共同inductiveARC protocol下分别固定VARC/TRM架构，把task variant数与total sample presentations分开控制；TTT只用每testtask demonstrations、queryinput做feedforward并inverse-transform/vote。; 5 pretrain×4 TTT budgets两模型40 pass@2结果；VARC scratch竞争力、TRM更依赖pretrain，TTTvariant数并非两模型都单调受益。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13660v1#S3", "https://arxiv.org/html/2610.13660v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13660v1", "sha256": "7c53ccd87967ed3507bb1b17e1889e507d07c31a134af8954850bad9f5b876b2" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13653", "title": "A Critical Look at Power Sampling when Test Time Sharpening LLMs", "published": "2026-10-09T14:29:28Z", "reviews": [ { "id": "2610.13653", "title": "A Critical Look at Power Sampling when Test Time Sharpening LLMs", "published": "2026-10-09T14:29:28Z", "decision": "reviewed_deferred", "reason": "Power sampling practical MCMC/SMC不收敛对照Best-ofN，并用Power-RLOO直接拟合target探索目标价值；MATH500/GPQA模型特异噪声结果，下一批需真正target sampling/梯度、不能用BoN冒称power distribution。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 The Pitfalls of Existing Sharpening Schemes", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13653v1", "sha256": "65486a1deedbb259bf7c63d5f002da225a2402b2400b78c159019dbc67a4a4ae", "local_html_sha256": "65486a1deedbb259bf7c63d5f002da225a2402b2400b78c159019dbc67a4a4ae", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13653v1", "result": "retrieved", "sha256": "65486a1deedbb259bf7c63d5f002da225a2402b2400b78c159019dbc67a4a4ae" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.13647", "title": "A-SAW: The Agentic Software Architecture Workflow", "published": "2026-10-09T14:25:24Z", "reviews": [ { "id": "2610.13647", "title": "A-SAW: The Agentic Software Architecture Workflow", "published": "2026-10-09T14:25:24Z", "decision": "reviewed_deferred", "reason": "已读摘要、A-SAW方法及评测片段：工业架构文档工作流含typed catalog、C4和human review；自然研究45文档13匹配对及访谈，bundle confounding不能解释为单独agent因果提升。合格工程评测候选待下一批。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2. Background", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3. Our Approach", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13647v1", "sha256": "d7e054d315f66fae60fd69d6685de9b97a4322e0d9570a34d163d9f5adc22c1f", "local_html_sha256": "d7e054d315f66fae60fd69d6685de9b97a4322e0d9570a34d163d9f5adc22c1f", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13647v1", "result": "retrieved", "sha256": "d7e054d315f66fae60fd69d6685de9b97a4322e0d9570a34d163d9f5adc22c1f" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13636", "title": "Benchmarking Answer-Grounding Consistency in Remote Sensing Visual Question Answering", "published": "2026-10-09T14:15:50Z", "reviews": [ { "id": "2610.13636", "title": "Benchmarking Answer-Grounding Consistency in Remote Sensing Visual Question Answering", "published": "2026-10-09T14:15:50Z", "decision": "reviewed_deferred", "reason": "AESP归一VLM自己的answer并验证predictedgrounding覆盖objectbox center；exist/count/compare规则支持answer，reported ES_all/ES_correct/jointsuccess而不只IOU。; 15027 G-RSVQA七releasedVLM，无新增训练；answer extractor三选一400响应人核，groundingprompt不含以前生成answer。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13636v1#S3", "https://arxiv.org/html/2610.13636v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13636v1", "sha256": "8a18c9ad61f8c143cf65e27b2ca96166f8c145093a5a271e63936781b3750638" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13635", "title": "PANDA: Prior-guided Attentional Dual-path Architecture", "published": "2026-10-09T14:15:14Z", "reviews": [ { "id": "2610.13635", "title": "PANDA: Prior-guided Attentional Dual-path Architecture", "published": "2026-10-09T14:15:14Z", "decision": "rejected_or_out_of_scope", "reason": "两阶段node→hyperedge→node HNN里main path插值receiver-normalizedattention与structuralincidence prior，auxiliary每sender sigmoid noncompetitive变换；还分析directedcomplexspectraloperator。; 四HNN/八nodeclassification datasets，normalisedaux同参消融含少数小例外。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13635v1#S3", "https://arxiv.org/html/2610.13635v1#S4", "https://arxiv.org/html/2610.13635v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13635v1", "sha256": "fb54054a6c23653fe4cb6a5ce3ffb42a25a72e1b00ff202954a00d04fb839169" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.13632", "title": "Humanize Olympiad Agents: A Multi-Session Workflow for Olympiad Problem Solving with Solve, Review, Wiki, Verify", "published": "2026-10-09T14:14:14Z", "reviews": [ { "id": "2610.13632", "title": "Humanize Olympiad Agents: A Multi-Session Workflow for Olympiad Problem Solving with Solve, Review, Wiki, Verify", "published": "2026-10-09T14:14:14Z", "decision": "reviewed_deferred", "reason": "已读摘要及HOA方法/评测片段：worker、独立reviewer、wiki检索与可信题面Lean验证协作；须分离语义有效性和kernel通过率、禁止被评agent读取gold证明，下一批。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13632v1", "sha256": "92bf11432f4a14abdb35bfaa102163781558fb568f7cf610c2426284b9fc4dec", "local_html_sha256": "92bf11432f4a14abdb35bfaa102163781558fb568f7cf610c2426284b9fc4dec", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13632v1", "result": "retrieved", "sha256": "92bf11432f4a14abdb35bfaa102163781558fb568f7cf610c2426284b9fc4dec" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13630", "title": "Multi-Stream Attention Residuals", "published": "2026-10-09T14:13:02Z", "reviews": [ { "id": "2610.13630", "title": "Multi-Stream Attention Residuals", "published": "2026-10-09T14:13:02Z", "decision": "accepted", "reason": "multi-head depth attention不沿hidden channel简单分块；各stream由多尺度因果卷积(4/8/12等)形成contextualviews，pseudo-query对历史单流sources作深度softmax，sigmoid lightweightmix合成单流layerinput。BlockMSAR只持久存single-streamsources，按需展开；最后streammerge与block边界按原方法保留。; 原论文18B/28B MoE十二downstreambench、iso-loss compute scaling与参数/compute/memory成本ablation；source reported1.58x不是本地实测速度。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "实施阶段已读摘要、作者信息、定义算法方法与主实验，以及列明的相关附录；复用实际实施审查，未宣称全部附录/整篇全文已读。", "review_artifact": "research-audits/oct12-review-closure/frozen-foundation-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13630v1", "sha256": "4322608f8a599ce171181dc5e4a667977274cf8d1795b12221189511b24fb5dc" }, "new_implementation_inferred_from_review": false } ], "discovery_origin": "independent-targeted-supplement", "implementation_status": "registered_scoped_runtime; actual acceptance tracked separately", "status": "accepted", "implementation_key": "msar", "accepted_scope": "bounded faithful mechanism/runtime or evaluation infrastructure; not original capability replication" }, { "id": "2610.13617", "title": "Training User Simulators with Latent Intrinsic Feedback", "published": "2026-10-09T14:00:36Z", "reviews": [ { "id": "2610.13617", "title": "Training User Simulators with Latent Intrinsic Feedback", "published": "2026-10-09T14:00:36Z", "decision": "reviewed_deferred", "reason": "官方PDF第一页确认Google Research/Georgia Tech；LES user simulator以real logs监督thought/rating/reaction，Sec4独立4000对话反馈/realism，Sec5实际multiturn assistant RL对比intrinsic/external rewards。Google优先合格候选，冻结16外defer下一批；须真实simulator与assistant training，不能伪造user反馈。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 User Simulators with Latent Evaluation States", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Evaluating the Feedback Behavior of User Simulators", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Training LLM Assistants with Simulators via Multi-Turn RL", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13617v1", "sha256": "e5fdc46eca3a6e95b37bba0e6e2fc249c995775dc9391357d9005946c6f095af", "local_html_sha256": "e5fdc46eca3a6e95b37bba0e6e2fc249c995775dc9391357d9005946c6f095af", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13617v1", "result": "retrieved", "sha256": "e5fdc46eca3a6e95b37bba0e6e2fc249c995775dc9391357d9005946c6f095af" } ], "alternate_evidence": { "url": "https://arxiv.org/pdf/2610.13617v1", "sha256": "f831e52cb9bbb95884277533bcae71cb2dabc47eb598cbc78501d7100c9efbf3", "location": "PDF page 1 author affiliations", "read_scope": "author/institution block only; method/evaluation sections read from official HTML" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.13605", "title": "zip2zip++: Towards Inference-Time Token Compression with Minimal Quality Loss", "published": "2026-10-09T13:52:54Z", "reviews": [ { "id": "2610.13605", "title": "zip2zip++: Towards Inference-Time Token Compression with Minimal Quality Loss", "published": "2026-10-09T13:52:54Z", "decision": "reviewed_deferred", "reason": "LZW动态hypertokens，untied in/out hyperencoders、identity residual初始化、size-aware RoPE、protected digit/special tokens、fault-tolerant guarded insertion及exact多segmentation概率。; §5四公开instruction模型与matchedLoRA预算，多corpus byte/word likelihood、math/code/IFEval/RULER与cumulative ablations；作者训练代码repo可访问。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract/author information/main defining methods/main experiments and tables; production-evidence paragraph spot check; appendices not exhaustively read", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13605v1#S3", "https://arxiv.org/html/2610.13605v1#S4", "https://arxiv.org/html/2610.13605v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13605v1", "sha256": "daf1a80ee646a2b6d7b277efb16fa77137677d355e518bacc05bf7381a4717cc" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13594", "title": "MENTOR: Learning from Failure with On-Policy Pedagogical Intervention", "published": "2026-10-09T13:43:49Z", "reviews": [ { "id": "2610.13594", "title": "MENTOR: Learning from Failure with On-Policy Pedagogical Intervention", "published": "2026-10-09T13:43:49Z", "decision": "reviewed_deferred", "reason": "已读§3方法及§4评测：MENTOR在全零GRPO组时用训练gold归因失败并生成/修复无泄漏hint，重采样后保持native GRPO；gold只能进训练teacher，推理不能读。DeepMath/LCB训练与独立数学/代码评测，78.3相对提升不是vs GRPO；下一批训练算法。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 12674 ] }, { "heading": "4 Experiments", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 8745 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13594v1", "sha256": "f25ba070bd3aa82ca28621c1719d4b01b70b3ddea234af01564cb0d03d5ba664", "local_html_sha256": "f25ba070bd3aa82ca28621c1719d4b01b70b3ddea234af01564cb0d03d5ba664", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13594v1", "result": "retrieved", "sha256": "f25ba070bd3aa82ca28621c1719d4b01b70b3ddea234af01564cb0d03d5ba664" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13578", "title": "RARA: Robust Policy Fine-Tuning through Retrieval-Anchored Representation Alignment for Robotic Manipulation", "published": "2026-10-09T13:31:20Z", "reviews": [ { "id": "2610.13578", "title": "RARA: Robust Policy Fine-Tuning through Retrieval-Anchored Representation Alignment for Robotic Manipulation", "published": "2026-10-09T13:31:20Z", "decision": "reviewed_deferred", "reason": "RARA从pretrain demos按taskintent/motion retrieval，stage1 target-representationcontrastivealignment到frozenpretrainfeature，stage2 BC finetune同时anchor retrieved representations；anchorloss不需要actionlabels。; DiffusionPolicy/GR00T、RoboTwin/RoboCasa/realrobot13tasks ID/OOD；额外对retrieved demos行为克隆反而降成功。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13578v1#S3", "https://arxiv.org/html/2610.13578v1#S4", "https://arxiv.org/html/2610.13578v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13578v1", "sha256": "03cf8c75f9004e0c4ea694cc7a54bc1e7203af929c6846a03f8b48259332203e" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13563", "title": "Not All Correct Fixes Run Alike: An Empirical Study of Patch Performance in Repository-Level Issue Resolution", "published": "2026-10-09T13:17:53Z", "reviews": [ { "id": "2610.13563", "title": "Not All Correct Fixes Run Alike: An Empirical Study of Patch Performance in Repository-Level Issue Resolution", "published": "2026-10-09T13:17:53Z", "decision": "reviewed_deferred", "reason": "将每个test-passing patch的官方F2P测试改为warmup+N次内部循环、G次独立进程；固定CPU/NUMA隔离，指令数和walltime以Welch ANOVA区分性能。; SWE-bench verified与Live过滤可测、至少两种不同correct solution的262案例；developer gold仅评价参照，不能给修复agent。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13563v1#S2", "https://arxiv.org/html/2610.13563v1#S3", "https://arxiv.org/html/2610.13563v1#S4", "https://arxiv.org/html/2610.13563v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13563v1", "sha256": "d22c56f27d08185f762cfefc035eb0bf42024a61a86a17be1b05f9e1ec4fdd36" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13557", "title": "Probabilistic Ship Trajectory Forecasting using Transformer Models on AIS Data", "published": "2026-10-09T13:13:07Z", "reviews": [ { "id": "2610.13557", "title": "Probabilistic Ship Trajectory Forecasting using Transformer Models on AIS Data", "published": "2026-10-09T13:13:07Z", "decision": "rejected_or_out_of_scope", "reason": "任务专用Transformer/Performer预测AIS位置quantiles；区域KDE历史heatmap经CNN进入网络，pinballloss输出0.025/0.5/0.975，10%训练holdout选收敛。; 月度1636220 vesseltrajectories与ARPA样constant/linear/Gaussianbaseline，heatmap帮助anticipate地理turn。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13557v1#S3", "https://arxiv.org/html/2610.13557v1#S4", "https://arxiv.org/html/2610.13557v1#S5", "https://arxiv.org/html/2610.13557v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13557v1", "sha256": "8fd9d9e988ae5211936c8144637c69866fc4926fc7451134f100b9215df3f845" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.13548", "title": "Auditing VLM Benchmarks from Existing Evaluation Runs", "published": "2026-10-09T13:05:22Z", "reviews": [ { "id": "2610.13548", "title": "Auditing VLM Benchmarks from Existing Evaluation Runs", "published": "2026-10-09T13:05:22Z", "decision": "reviewed_deferred", "reason": "VLMEvalKit保存预测/判分按questionid对齐，新增no-image blindrun，六postprocessdetectors+matched-permutationnull/interval/modelpool敏感性；gold只外部audit。; MMStar三模型和binary format calibration边界；曾goldbackfill把unanimousconflicts134压成12，已拒绝这种补值。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13548v1#S3", "https://arxiv.org/html/2610.13548v1#S4", "https://arxiv.org/html/2610.13548v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13548v1", "sha256": "54bedf686d50e27a471011fc3b36f81d9e74c81b922bcefdf4ff0875c3cc96fd" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13527", "title": "S-Space: Exploring Spatial Workspace in Multimodal Language Models", "published": "2026-10-09T12:48:41Z", "reviews": [ { "id": "2610.13527", "title": "S-Space: Exploring Spatial Workspace in Multimodal Language Models", "published": "2026-10-09T12:48:41Z", "decision": "reviewed_deferred", "reason": "objecttoken activations的spatialanswer-score gradients识别水平/垂直/距离三轴linearS-Space，readout相对坐标margin，并用activation intervention检验verbal spatial因果使用。; COCO轴识别、CV-Bench/EmbSpatial/SpatialTunnel不同prompts，MMSI viewpoint/reference-frame失败分解。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13527v1#S2", "https://arxiv.org/html/2610.13527v1#S3", "https://arxiv.org/html/2610.13527v1#S4", "https://arxiv.org/html/2610.13527v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13527v1", "sha256": "bbbbddef4d8d12075ae6f05e721de39297c1f69061ce6c576b67fb4264a7e77c" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13521", "title": "Beyond Domain-Routed Supervision: Cross-Teacher Guarding for Multi-Teacher On-Policy Distillation", "published": "2026-10-09T12:41:51Z", "reviews": [ { "id": "2610.13521", "title": "Beyond Domain-Routed Supervision: Cross-Teacher Guarding for Multi-Teacher On-Policy Distillation", "published": "2026-10-09T12:41:51Z", "decision": "accepted", "reason": "多教师 token credit 加同轨迹 preventive gradient projection 两部分均有原文方法与对照证据，纳入冻结批次；投影数学性质不充当生产/能力收益证明。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "official abstract; full PDF pages 3-8 plus first 4000 characters of page 9 method/evaluation/ablation content; not all references/appendices", "review_artifact": "research-audits/oct12-review-closure/frozen-post-training-reviews.json", "evidence_locations": [ { "kind": "method", "heading": "3.1 Problem Setup; 3.2 Cross-Teacher-Guided Credit Assignment", "pdf_pages": [ 3, 4, 5 ], "equations": [ 1, 2, 3, 4, 5 ], "figure": 2, "scope": "routed and non-routed teacher score the same student trajectory against a frozen reference; agreement modulates target token credit" }, { "kind": "method", "heading": "3.3 Preventive Multi-Teacher Gradient Projection", "pdf_pages": [ 5, 6 ], "equations": [ 6, 7, 8, 9 ], "scope": "actual auxiliary OPD gradient on identical stored target-domain trajectory; per-parameter-tensor conflict projection with norm-retention floor" }, { "kind": "method", "heading": "3.4 Overall Objective and Training", "pdf_pages": [ 6 ], "equations": [ 10, 11 ], "scope": "sequence mean then domain mean; only corrected target gradients are averaged/applied, auxiliary gradients are constraints rather than extra optimizer updates" }, { "kind": "evaluation", "heading": "4.1 Experimental Setup; 4.2 Main Results", "pdf_pages": [ 6, 7, 8 ], "tables": [ 1, 2, 3 ], "scope": "Code/Math/Medical public benchmarks, Avg@4/32/1 protocols, two student sizes, low-conflict and three-domain controls" }, { "kind": "evaluation", "heading": "4.3 Further Analysis", "pdf_pages": [ 8, 9 ], "table": 4, "figures": [ 3, 4 ], "page_9_characters_read": [ 0, 4000 ], "scope": "same-trajectory versus PCGrad compatibility, credit/projection/adaptive-retention ablations; local gradient statements are not capability guarantees" } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/pdf/2610.13521v1", "sha256": "464e401ddde486ea2ee908b07f4d359d4591e5136e2ba5f2dc02dbd7089443a6", "local_path": "tmp/pdfs/2610.13521.pdf", "local_pdf_sha256_matches": true, "pdf_pages": 30, "version_evidence": "PDF page 1 identifies arXiv:2610.13521v1 and 9 Oct 2026" }, "new_implementation_inferred_from_review": false } ], "discovery_origin": "independent-targeted-supplement", "implementation_status": "registered_scoped_runtime; actual acceptance tracked separately", "status": "accepted", "implementation_key": "road-mopd", "accepted_scope": "bounded faithful mechanism/runtime or evaluation infrastructure; not original capability replication" }, { "id": "2610.13510", "title": "Retrieval-Augmented Multimodal Language Model with Heterogeneous Evidence Integration for Clinical Prognosis", "published": "2026-10-09T12:35:22Z", "reviews": [ { "id": "2610.13510", "title": "Retrieval-Augmented Multimodal Language Model with Heterogeneous Evidence Integration for Clinical Prognosis", "published": "2026-10-09T12:35:22Z", "decision": "reviewed_deferred", "reason": "PRISM numericalfeaturetokens+ECGencoder+recencyweightedpriorECG与training-cohort dualmodalitypopulationmemory meanpooledretrieval embeddings，真实softtokens入LLM，不暴露individualretrievedrecords。; MDS-ED1443tasks和外部MC-MEDsupervisedadapt；96hprearrival输入为baseline没有的证据，benchmark-window matched结果不胜最强baseline。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13510v1#S3", "https://arxiv.org/html/2610.13510v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13510v1", "sha256": "4f810eb2996c78c9e3d62b815fdb79b48f2076c1dbb0442fdd9822ab141749d2" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13504", "title": "Beyond Temporal Proximity: Explanatory World Models for Credit Assignment in Reinforcement Learning", "published": "2026-10-09T12:30:14Z", "reviews": [ { "id": "2610.13504", "title": "Beyond Temporal Proximity: Explanatory World Models for Credit Assignment in Reinforcement Learning", "published": "2026-10-09T12:30:14Z", "decision": "reviewed_deferred", "reason": "Spotify EWM用解释图递归重新分配延迟奖励并保留未折扣episode return；Crafter/WordCraft可作后续通用RL候选。§5.2明确hidden termination rule模拟用户，非线上生产A/B；§5.3是held-out日志离线replay。工业收录门槛未过，通用后训练部分转下一批，非这16项实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "bounded_methods_and_evaluation", "review_artifact": "research-audits/oct12-review-closure/recommendation-reviews.json", "evidence_locations": [ "§3 Explanatory World Models", "§4.1–4.2", "§5.1–5.3" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13504v1", "sha256": "7afbcadb10735c1f26375b49759803de18826164027fab0e91b16d2085041258" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training", "recommendation" ], "status": "reviewed_deferred" }, { "id": "2610.13494", "title": "Humanize MLE Agents: Multi-Agent Session Alternation Beyond Single-Agent Limits", "published": "2026-10-09T12:26:34Z", "reviews": [ { "id": "2610.13494", "title": "Humanize MLE Agents: Multi-Agent Session Alternation Beyond Single-Agent Limits", "published": "2026-10-09T12:26:34Z", "decision": "reviewed_deferred", "reason": "已读摘要及CR-SMDP方法片段：HMA在共享workspace交替两个agent、session cap及上下文重置，报告75 MLE-bench任务；需以验证集选session cap并隔离任务反馈，进一步读完整实验再实现，下一批。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 Motivation: Limits of Single-Agent Exploration", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13494v1", "sha256": "e88a491680927a7ac348c75dfbdfeeb8c1f043bcdb5e99cc03c9bee6f813568c", "local_html_sha256": "e88a491680927a7ac348c75dfbdfeeb8c1f043bcdb5e99cc03c9bee6f813568c", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13494v1", "result": "retrieved", "sha256": "e88a491680927a7ac348c75dfbdfeeb8c1f043bcdb5e99cc03c9bee6f813568c" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13485", "title": "Proactive Agentic Orchestration", "published": "2026-10-09T12:18:49Z", "reviews": [ { "id": "2610.13485", "title": "Proactive Agentic Orchestration", "published": "2026-10-09T12:18:49Z", "decision": "reviewed_deferred", "reason": "POrchestra允许subagent主动申请harness变化并保留局部执行，OPOD从hindsight request回归orchestrator决定作过程监督；GAIA/动态数字助手/软件任务，转agent orchestrator下一批需真实控制与训练而非静态标签。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13485v1", "sha256": "9649e66ff2a76853ae90b9e13560b1e1271b96ee33f6a0e6289df5a16e850df3", "local_html_sha256": "9649e66ff2a76853ae90b9e13560b1e1271b96ee33f6a0e6289df5a16e850df3", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13485v1", "result": "retrieved", "sha256": "9649e66ff2a76853ae90b9e13560b1e1271b96ee33f6a0e6289df5a16e850df3" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.13481", "title": "Branching On-Policy Distillation", "published": "2026-10-09T12:13:34Z", "reviews": [ { "id": "2610.13481", "title": "Branching On-Policy Distillation", "published": "2026-10-09T12:13:34Z", "decision": "accepted", "reason": "符合冻结批次的后训练/蒸馏核心算法，相关方法、对照与实际 token 成本已审；用户明确授权实施，不将离线 agent benchmark 冒称 production A/B。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "official abstract; PDF selected method/experiment pages and generation-cost appendix, not references or all appendices", "review_artifact": "research-audits/oct12-review-closure/frozen-post-training-reviews.json", "evidence_locations": [ { "kind": "method", "heading": "4.1 Selecting Fork Points; 4.2 Conditional Resampling and Coverage", "pdf_pages": [ 4, 5 ], "equations": [ 6, 7, 8 ], "scope": "PDF pages 4 and 5 read in full; student uncertainty/disagreement ranking, frozen student-policy suffix sampling" }, { "kind": "method", "heading": "4.3 Training Objective", "pdf_pages": [ 5 ], "equations": [ 9, 10, 11 ], "scope": "shared-prefix mask, token-mean PPO OPD objective; not teacher-forced branches or reward-selected suffixes" }, { "kind": "method", "heading": "4.4 Protocol-Boundary Repair", "pdf_pages": [ 6 ], "equations": [ 12, 13 ], "characters_read": [ 0, 3500 ], "scope": "positive/negative protocol-event supervision and task-specific detector; empty constraints mean zero auxiliary loss" }, { "kind": "evaluation", "heading": "5.1 Experimental Setup; 5.2 Main Results", "pdf_pages": [ 6, 7 ], "tables": [ 1, 2 ], "scope": "page 6 first 3500 characters and full page 7; mathematics avg@16 and single-trajectory agent evaluation, seven-rollout/PBR controls" }, { "kind": "evaluation", "heading": "B.4 Generation and Environment Costs", "pdf_pages": [ 22 ], "tables": [ 11, 12 ], "scope": "full PDF page 22; measured tokens distinguished from position estimates and extra environment replay cost" } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/pdf/2610.13481v1", "sha256": "501993b6c44418347f45a2cd280cf987308bb525dc1a90ee90855f0ffa61a042", "local_path": "tmp/pdfs/2610.13481.pdf", "local_pdf_sha256_matches": true, "pdf_pages": 33, "version_evidence": "PDF page 1 identifies arXiv:2610.13481v1 and 9 Oct 2026" }, "new_implementation_inferred_from_review": false } ], "discovery_origin": "independent-targeted-supplement", "implementation_status": "registered_scoped_runtime; actual acceptance tracked separately", "status": "accepted", "implementation_key": "branch-opd", "accepted_scope": "bounded faithful mechanism/runtime or evaluation infrastructure; not original capability replication" }, { "id": "2610.13455", "title": "Agentic Large Language Models", "published": "2026-10-09T11:57:04Z", "reviews": [ { "id": "2610.13455", "title": "Agentic Large Language Models", "published": "2026-10-09T11:57:04Z", "decision": "rejected_or_out_of_scope", "reason": "已读摘要及定义/教材目的段：研究生课程AgenticLLMs教材综合现有概念，不是可单独复现的新算法或基准，作为背景参考。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "1.1 Why this book?", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "1.2 Concepts and Definitions", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13455v1", "sha256": "666575c8adada2fff6f9321826a1e48490deb4880abe27ff4fd3160d25db6580", "local_html_sha256": "666575c8adada2fff6f9321826a1e48490deb4880abe27ff4fd3160d25db6580", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13455v1", "result": "retrieved", "sha256": "666575c8adada2fff6f9321826a1e48490deb4880abe27ff4fd3160d25db6580" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.13439", "title": "Collective Autonomous Vehicle Routing Needs Social Awareness for Coexistence with Human Drivers", "published": "2026-10-09T11:37:47Z", "reviews": [ { "id": "2610.13439", "title": "Collective Autonomous Vehicle Routing Needs Social Awareness for Coexistence with Human Drivers", "published": "2026-10-09T11:37:47Z", "decision": "rejected_or_out_of_scope", "reason": "已读摘要、前置与方法段：RouteRL/SUMO中用反事实边际社会成本训练自动车协同路由，是交通MARL而非LLM agent；真实路网拓扑不能当真实线上流量A/B。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2  Preliminaries", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3  Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13439v1", "sha256": "81cf4dc18b38fe7c2b7d0c9d6f32be7e509336044299fb49042211c536293717", "local_html_sha256": "81cf4dc18b38fe7c2b7d0c9d6f32be7e509336044299fb49042211c536293717", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13439v1", "result": "retrieved", "sha256": "81cf4dc18b38fe7c2b7d0c9d6f32be7e509336044299fb49042211c536293717" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.13438", "title": "Rollout-Time Intervention: Unlocking Agentic Reinforcement Learning Beyond Policy Limits", "published": "2026-10-09T11:37:09Z", "reviews": [ { "id": "2610.13438", "title": "Rollout-Time Intervention: Unlocking Agentic Reinforcement Learning Beyond Policy Limits", "published": "2026-10-09T11:37:09Z", "decision": "reviewed_deferred", "reason": "已读摘要及RTI/实验片段：训练时reward-side干预、perplexity gate与reasoning token权重，使干预trajectory可以参与离线更新；需真训练及privileged information训练/测试隔离，下一批后训练候选。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Preliminaries", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Rollout-Time Intervention", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13438v1", "sha256": "0c7472ea48d0ef38fde12aa730ede8d0ee87115fd04838b2ac5a6e3e8b90f24e", "local_html_sha256": "0c7472ea48d0ef38fde12aa730ede8d0ee87115fd04838b2ac5a6e3e8b90f24e", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13438v1", "result": "retrieved", "sha256": "0c7472ea48d0ef38fde12aa730ede8d0ee87115fd04838b2ac5a6e3e8b90f24e" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13421", "title": "REINS: Agent-guided Thread Scheduling for Concurrency Bug Detection and Reproduction", "published": "2026-10-09T11:27:12Z", "reviews": [ { "id": "2610.13421", "title": "REINS: Agent-guided Thread Scheduling for Concurrency Bug Detection and Reproduction", "published": "2026-10-09T11:27:12Z", "decision": "reviewed_deferred", "reason": "已读摘要及REINS概览/评测片段：LLM提出语义并发假设，调度器施加order/state约束后真实试验反驳/确认；公开bench和新bug作者确认不是生产A/B。后续须执行真实程序而非静态fixture。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 Overview", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Preliminary", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13421v1", "sha256": "81a0143ff5bbe93c964e60cc181760e1dd9b40ed7aa3ffaee3ccf9ad89d87849", "local_html_sha256": "81a0143ff5bbe93c964e60cc181760e1dd9b40ed7aa3ffaee3ccf9ad89d87849", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13421v1", "result": "retrieved", "sha256": "81a0143ff5bbe93c964e60cc181760e1dd9b40ed7aa3ffaee3ccf9ad89d87849" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13416", "title": "Web$hell: A Unified Interface for Agentic Web Access", "published": "2026-10-09T11:23:40Z", "reviews": [ { "id": "2610.13416", "title": "Web$hell: A Unified Interface for Agentic Web Access", "published": "2026-10-09T11:23:40Z", "decision": "reviewed_deferred", "reason": "已读摘要及Web$hell实验片段：typed WebContract、WebCompiler与WebCache将网页操作转成可地址信息/行动；检索/导航/深研究公开评测适合后续基础设施核验，本次不扩scope。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Preliminary Analysis: Can agents effectively access the web through existing access models?", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Web$hell", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13416v1", "sha256": "e2b5257bca56043c841990bebbfa0604aeb711c60e2079f7d3ac76e829c2f0bd", "local_html_sha256": "e2b5257bca56043c841990bebbfa0604aeb711c60e2079f7d3ac76e829c2f0bd", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13416v1", "result": "retrieved", "sha256": "e2b5257bca56043c841990bebbfa0604aeb711c60e2079f7d3ac76e829c2f0bd" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13411", "title": "In-Context Learning for System Identification: Probabilistic Formulation, Long-Sequence Handling, and Benchmark Validation", "published": "2026-10-09T11:21:14Z", "reviews": [ { "id": "2610.13411", "title": "In-Context Learning for System Identification: Probabilistic Formulation, Long-Sequence Handling, and Benchmark Validation", "published": "2026-10-09T11:21:14Z", "decision": "reviewed_deferred", "reason": "synthetic systemclass预训probabilisticTransformer encoder-decoder metamodel，recurrentpatch neuralnetwork处理长context；inference多窗inversevariance合成，uncertaintyaware合成数据可regularize传统parametricID。; Wiener-Hammerstein syntheticpretrain与realnonlinearstandardbench/ OOD input-onlyadapt，RTX4090；部分传统模型RMSE略更优。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13411v1#S3", "https://arxiv.org/html/2610.13411v1#S4", "https://arxiv.org/html/2610.13411v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13411v1", "sha256": "db3342fbf4f2bb2b17cdc4a8a7d674a2d57320ebe0d100e35bef857c2104c198" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13398", "title": "Towards the Roof of the World: NLP Progress in Tibetan Languages", "published": "2026-10-09T11:11:10Z", "reviews": [ { "id": "2610.13398", "title": "Towards the Roof of the World: NLP Progress in Tibetan Languages", "published": "2026-10-09T11:11:10Z", "decision": "rejected_or_out_of_scope", "reason": "藏语NLP系统文献计量综述140篇，不是新可执行训练/Agent算法或独立新benchmark。本次读完整摘要定界，不假称方法全文已审。", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "read_scope": "abstract_scope_triage", "review_artifact": "research-audits/oct12-review-closure/recommendation-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13398v1", "sha256": "82ae30a24931990d2cc495c767a9cc28c99fd268881125f3c223b0e12f99815d" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "recommendation" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.13395", "title": "ProgramGym: Scaling Training Environments for Long-Horizon Program Reproduction Agents", "published": "2026-10-09T11:09:54Z", "reviews": [ { "id": "2610.13395", "title": "ProgramGym: Scaling Training Environments for Long-Horizon Program Reproduction Agents", "published": "2026-10-09T11:09:54Z", "decision": "reviewed_deferred", "reason": "已读ProgramEnv、ProgramAgent、评测段：1648多语言程序环境以隔离的gold执行生成黑盒测试，pass/double/dummy/弱断言验证；agent角色分离、固定revision、独立compile/recovery，不能让implementer看gold源。作者报告SFT/scaffold分层结果，resolved近零不能宣传全能力复现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Program-Env", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 5484 ] }, { "heading": "5 Program-Agent", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 4017 ] }, { "heading": "6 Experiments and Analysis", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 8864 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13395v1", "sha256": "af4313ae6d153bb7a037dec93558caaa676b180672b2f94bd41b73bf773fce45", "local_html_sha256": "af4313ae6d153bb7a037dec93558caaa676b180672b2f94bd41b73bf773fce45", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13395v1", "result": "retrieved", "sha256": "af4313ae6d153bb7a037dec93558caaa676b180672b2f94bd41b73bf773fce45" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13370", "title": "Overcoming Data Imbalances with Expert Selection Biases", "published": "2026-10-09T10:45:19Z", "reviews": [ { "id": "2610.13370", "title": "Overcoming Data Imbalances with Expert Selection Biases", "published": "2026-10-09T10:45:19Z", "decision": "reviewed_deferred", "reason": "COAX按输入category给top-K expertselection加bias，使少数category偏向可过配容量的expert subset同时可跨组共享；改变池子容量不改变每tokenK=8 activecompute。; 多模态/多语/domain三MoE规模到10.6B，分配/alphasweep，CPT新语言Russian/Hindi略弱于vanilla但replay更好。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13370v1#S3", "https://arxiv.org/html/2610.13370v1#S4", "https://arxiv.org/html/2610.13370v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13370v1", "sha256": "878af635673c69d2e9fd60a9e766d752e255033ad3b64e5c59934a41a904376a" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13361", "title": "AgentSWE: Can Coding Agents Build the Agent You Actually Want?", "published": "2026-10-09T10:35:27Z", "reviews": [ { "id": "2610.13361", "title": "AgentSWE: Can Coding Agents Build the Agent You Actually Want?", "published": "2026-10-09T10:35:27Z", "decision": "reviewed_deferred", "reason": "已读AgentSWE benchmark与实验缓存片段：评估实际交付agent的构建/编辑/优化任务，public dev与heldout workspace隔离、rubric/integrity gates。25任务、heldout很小，不能用静态代码或routing答案替代交付评测；下一批infra。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 The AgentSWE Benchmark", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 6406 ] }, { "heading": "5 Experiments", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 7500 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13361v1", "sha256": "71950e44731ccf681842eb42e7812062fefd08a96329672495ac280b872eb5a8", "local_html_sha256": "71950e44731ccf681842eb42e7812062fefd08a96329672495ac280b872eb5a8", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13361v1", "result": "retrieved", "sha256": "71950e44731ccf681842eb42e7812062fefd08a96329672495ac280b872eb5a8" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13356", "title": "SWE-Critix: Post-Training Generalist Verifiers for Assessing Coding-Agent Trajectories", "published": "2026-10-09T10:30:21Z", "reviews": [ { "id": "2610.13356", "title": "SWE-Critix: Post-Training Generalist Verifiers for Assessing Coding-Agent Trajectories", "published": "2026-10-09T10:30:21Z", "decision": "reviewed_deferred", "reason": "已读SWE-Critix方法与实验：成败成对teacher CoT生成pointwise SFT，然后实际GRPO二元label reward+KL；issue级20%隔离，负例召回单独报告。43k评测与128 Ascend成本，后续必须训练真实verifier，非常数fitness。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "II SWE-Critix", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 8887 ] }, { "heading": "III Preliminary Evaluation", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 2351 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13356v1", "sha256": "1f9ea2d7dff47aec4f9530da617f8e0b2cea0abfcdd9d3af8ea7c259c828c38a", "local_html_sha256": "1f9ea2d7dff47aec4f9530da617f8e0b2cea0abfcdd9d3af8ea7c259c828c38a", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13356v1", "result": "retrieved", "sha256": "1f9ea2d7dff47aec4f9530da617f8e0b2cea0abfcdd9d3af8ea7c259c828c38a" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13348", "title": "Test-Time Memory Construction", "published": "2026-10-09T10:24:51Z", "reviews": [ { "id": "2610.13348", "title": "Test-Time Memory Construction", "published": "2026-10-09T10:24:51Z", "decision": "accepted", "reason": "用户点名冻结16范围；selected defining-method/experiment原文证据核验，具体实现/运行等级独立记录。; 离线query-agnostic metadata保raw links；当前query驱动test-time progressive memory cache，自主SearchMeta/ReadRaw/Answer，去重且只读已观察sourceID。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "实施期间读取摘要、中心方法/训练/评测及部分相关附录；本次复用这些已读来源并重新校验官方HTML字节SHA和v1日期，不声明全文每段已读。", "review_artifact": "research-audits/oct12-review-closure/frozen-agent-reviews.json", "evidence_locations": [ "3.1 Disentangling Memory Orchestration from Offline to Test Time", "3.2 Dynamic Memory Construction During LLM Reasoning", "4 Experimental Methodology", "5.1 Overall Performance", "5.2 Ablation Studies", "5.3 Effectiveness of Test-Time Computation", "A.3 Additional Experimental Details", "A.8 Prompt Templates" ], "dateproof": { "url": "https://arxiv.org/abs/2610.13348", "sha256": "3f46e16b1a17a8bc92c18bac3e65deb137589ef17c78f1e1d96967c90111b9a3", "location": "Submission history / [v1]", "first_publication_utc": "2026-10-09T10:24:51Z" }, "source_proof": { "url": "https://arxiv.org/html/2610.13348v1", "sha256": "41ff56ee7707b7c75d6925023e706775d9bde4e3424f932af1f9b6427085b484", "byte_hash_verified_on": "2026-10-12" }, "new_implementation_inferred_from_review": false } ], "discovery_origin": "independent-targeted-supplement", "implementation_status": "registered_scoped_runtime; actual acceptance tracked separately", "status": "accepted", "implementation_key": "t2mem", "accepted_scope": "bounded faithful mechanism/runtime or evaluation infrastructure; not original capability replication" }, { "id": "2610.13341", "title": "Understanding Mamba-3 Through Oscillatory Dynamics", "published": "2026-10-09T10:16:58Z", "reviews": [ { "id": "2610.13341", "title": "Understanding Mamba-3 Through Oscillatory Dynamics", "published": "2026-10-09T10:16:58Z", "decision": "reviewed_deferred", "reason": "Mamba3 decay×inputdependentrotation的statepair等价underdampedsecondorderdynamics；控制相同state结构比critical/overdamped，保持读写phase一致并做pretrainedphaseintervention。; state tracking、1.5BMamba3/SISO及语言pretrain长context/lengthgeneralization；learnedphase去除扰动退化。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13341v1#S3", "https://arxiv.org/html/2610.13341v1#S4", "https://arxiv.org/html/2610.13341v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13341v1", "sha256": "0f61fb1aa043fc0f13193846c7be123aba91aefd9e2c90b41d8c4df5dfbbe4d2" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13337", "title": "Exploring Attention Sinks in Text-to-Image Transformers from Diffusion and Autoregressive Models", "published": "2026-10-09T10:10:51Z", "reviews": [ { "id": "2610.13337", "title": "Exploring Attention Sinks in Text-to-Image Transformers from Diffusion and Autoregressive Models", "published": "2026-10-09T10:10:51Z", "decision": "reviewed_deferred", "reason": "按prompt/step attention cumulative mass定位T2I head/positionsinks，在FLUX/ZImage/Infinity上logit/value/head干预，并插registertokens再分配sinks。; GenEval/DPG、HPSv2/ImageReward，H100；diffusion边界/corner vsAR首token/global，head稳定而positions随prompt变。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13337v1#S3", "https://arxiv.org/html/2610.13337v1#S4", "https://arxiv.org/html/2610.13337v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13337v1", "sha256": "ee9c38c3ac95b3e899334b27bf7eb63f509cdf5a1359656478f5d6c57ff4248d" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13335", "title": "AgentTune: A Unified Training Lifecycle for Tool-Using LLM Agents", "published": "2026-10-09T10:09:27Z", "reviews": [ { "id": "2610.13335", "title": "AgentTune: A Unified Training Lifecycle for Tool-Using LLM Agents", "published": "2026-10-09T10:09:27Z", "decision": "reviewed_deferred", "reason": "原文缓存后来可用，补读安装与library structure片段：AgentTune统一EventLog、RL trainer及DECIDE，LSAL source-available非无限制开源，依赖TRL签名版本。作者repo尚未独立审计，部分自愈仅设计；下一批核验真实训练/运行而非宣称已集成。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2  Installation and Getting Started", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3  Library Structure", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13335v1", "sha256": "17bdb0d89b79fe2c674624e83fe7ad153b5c6e709e3006901de9fbe6a10933b3", "local_html_sha256": "17bdb0d89b79fe2c674624e83fe7ad153b5c6e709e3006901de9fbe6a10933b3", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13335v1", "result": "retrieved", "sha256": "17bdb0d89b79fe2c674624e83fe7ad153b5c6e709e3006901de9fbe6a10933b3" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13334", "title": "PDE-PFN: Prior-Data Fitted Neural PDE Solver", "published": "2026-10-09T10:08:51Z", "reviews": [ { "id": "2610.13334", "title": "PDE-PFN: Prior-Data Fitted Neural PDE Solver", "published": "2026-10-09T10:08:51Z", "decision": "reviewed_deferred", "reason": "PDE-PFN以noisyapproxPINN solutions为prior，self/crossattention PFN对contextcoord/value与querycoord预测posteriorpredictivemean，训练含AGFreg并按contextsize缩放。; 2DCDR family无taskfinetune测试和SWE/CNSE crossPDE fine-tuning；DPOT原一步任务改四步影响baseline。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13334v1#S3", "https://arxiv.org/html/2610.13334v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13334v1", "sha256": "ca725de41abe06938307377afc2a20339d43eb2fd52b4bd25606b5cc9d14cc73" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13330", "title": "Masked Diffusion Language Agents for Tool-Integrated Chemical Reasoning", "published": "2026-10-09T10:07:37Z", "reviews": [ { "id": "2610.13330", "title": "Masked Diffusion Language Agents for Tool-Integrated Chemical Reasoning", "published": "2026-10-09T10:07:37Z", "decision": "reviewed_deferred", "reason": "ChemDiffAgent 对交织tool trajectory做masked denoising，agentic SFT+variance-reduced preference optimization；6类化学任务有匹配预算评测，核心需要实际diffusion backbone和化学tool graph，不能普通AR proxy。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "4 ChemDiffAgent", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13330v1", "sha256": "2a9cbaa7ce1108b2cf91a4a5fa3c4a7bf90af533a17fa9637a86950d224dd05b", "local_html_sha256": "2a9cbaa7ce1108b2cf91a4a5fa3c4a7bf90af533a17fa9637a86950d224dd05b", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13330v1", "result": "retrieved", "sha256": "2a9cbaa7ce1108b2cf91a4a5fa3c4a7bf90af533a17fa9637a86950d224dd05b" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.13325", "title": "Taking Tokenization off the Host for Agentic LLM Serving", "published": "2026-10-09T10:05:09Z", "reviews": [ { "id": "2610.13325", "title": "Taking Tokenization off the Host for Agentic LLM Serving", "published": "2026-10-09T10:05:09Z", "decision": "reviewed_deferred", "reason": "已读摘要及GraphTok设计/评测片段：CUDA graph的GPU精确BPE加CPU短串dispatch，5 tokenizer bit-exact与服务latency/throughput，非生产A/B。下一批serving/foundation候选须真实GPU与分派exactness验证。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 Motivation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Background", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13325v1", "sha256": "b683906dafd48ef9ea2320204325ecc4837e4d2eeb90208a12f1703d400009a2", "local_html_sha256": "b683906dafd48ef9ea2320204325ecc4837e4d2eeb90208a12f1703d400009a2", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13325v1", "result": "retrieved", "sha256": "b683906dafd48ef9ea2320204325ecc4837e4d2eeb90208a12f1703d400009a2" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13321", "title": "Host-Aware Web Crawling for LLM Pretraining Data", "published": "2026-10-09T10:03:04Z", "reviews": [ { "id": "2610.13321", "title": "Host-Aware Web Crawling for LLM Pretraining Data", "published": "2026-10-09T10:03:04Z", "decision": "reviewed_deferred", "reason": "从WARC质量classifier/hostgraph/coverage估计host间ICC，offline模拟先sample少数pages、按hostsamplequality决定crawl depth，而非只用harmoniccentrality分预算。; CC2024-10约0.25%WARC classifiedsample与另fullcrawl预计算hostartifact两种regime；centrality某classifier不如random。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13321v1#S3", "https://arxiv.org/html/2610.13321v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13321v1", "sha256": "7ff8201e10614e1ef87a893aa92e6edd3e1d2b0edc3a702b59f8296bc72eddc0" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13296", "title": "PRIME: Pre-Retrieval Index Modeling Enhancement for Structured Document Retrieval-Real", "published": "2026-10-09T09:44:47Z", "reviews": [ { "id": "2610.13296", "title": "PRIME: Pre-Retrieval Index Modeling Enhancement for Structured Document Retrieval-Real", "published": "2026-10-09T09:44:47Z", "decision": "reviewed_deferred", "reason": "已读摘要与PRIME方法/实验片段：corpus-side语义重构、确定性metadata与层级检索，在QASPER/BRIGHT/NFCorpus衡量quality-fidelity-cost；strong reranker可消除index收益，full deployment不等于生产A/B。下一批RAG/检索候选，需防重构失真。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13296v1", "sha256": "04858f69f420e6d61e97d57f05f7c11973d52de6103299e7fce202d0f06ce433", "local_html_sha256": "04858f69f420e6d61e97d57f05f7c11973d52de6103299e7fce202d0f06ce433", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13296v1", "result": "retrieved", "sha256": "04858f69f420e6d61e97d57f05f7c11973d52de6103299e7fce202d0f06ce433" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13285", "title": "BoundaryEvo: Data-Model Co-Evolution at the Capability Boundary for Long-Chain Reasoning Distillation", "published": "2026-10-09T09:37:33Z", "reviews": [ { "id": "2610.13285", "title": "BoundaryEvo: Data-Model Co-Evolution at the Capability Boundary for Long-Chain Reasoning Distillation", "published": "2026-10-09T09:37:33Z", "decision": "accepted", "reason": "能力 priming 后随学生变化而重估边界和训练分布是明确核心机制，相关方法/控制实验及原 prompt 契约已读；纳入冻结实施，不以接口/解析测试代替真实 co-evolution。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "official abstract; PDF pages 3-8 method/main experimental sections and F.1-F.3 raw prompt/threshold evidence on pages 26-27; not every appendix or reference", "review_artifact": "research-audits/oct12-review-closure/frozen-post-training-reviews.json", "evidence_locations": [ { "kind": "method", "heading": "3.1 Stage 1: Judge-Free Direct Distillation for Capability Priming", "pdf_pages": [ 3, 4 ], "equations": [ 1 ], "scope": "teacher behavior after basic validity checks, without correctness/capability-gap filtering" }, { "kind": "method", "heading": "3.2 Stage 2: Dynamic Capability-Boundary Co-Evolution", "pdf_pages": [ 4, 5 ], "equations": [ 2, 3, 4, 5, 6, 7, 8, 9 ], "figure": 2, "scope": "independent current-student/fixed-expert rollouts, analyzer quality/validity, rechecked replay plus novel acquisition, fresh multiple teacher trajectories then independent verification" }, { "kind": "method", "heading": "3.3 Training Objective", "pdf_pages": [ 5 ], "equations": [ 10 ], "scope": "response-token-normalized SFT after priming or verified boundary selection" }, { "kind": "evaluation", "heading": "4.1-4.6 Experiments", "pdf_pages": [ 6, 7, 8 ], "tables": [ 1, 2, 3 ], "scope": "same-priming/1K-budget random and LLM2LLM controls, three turns, 11 benchmarks; Qwen3.5-2B gains need not be monotonic" }, { "kind": "implementation_protocol", "heading": "F.1 Stage 1 Teacher Prompt; F.2 Student/Expert Prompt; F.3 Capability-Boundary Analyzer Prompt", "pdf_pages": [ 26, 27 ], "scope": "full selected pages: no gold to solvers, reference-aware external analyzer, strict JSON and score thresholds; reference-correctness/leakage hard rejection switches disabled in original reported route" } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/pdf/2610.13285v1", "sha256": "d5e5f36ff36642a787638f484af3f54a4d25cc25db85c5bbde082fb6c25230a0", "local_path": "tmp/pdfs/2610.13285.pdf", "local_pdf_sha256_matches": true, "pdf_pages": 29, "version_evidence": "PDF page 1 identifies arXiv:2610.13285v1 and 9 Oct 2026" }, "new_implementation_inferred_from_review": false } ], "discovery_origin": "independent-targeted-supplement", "implementation_status": "registered_scoped_runtime; actual acceptance tracked separately", "status": "accepted", "implementation_key": "boundary-evo", "accepted_scope": "bounded faithful mechanism/runtime or evaluation infrastructure; not original capability replication" }, { "id": "2610.13276", "title": "CopRoute: Dependence-Aware Conformal Routing for Language Models", "published": "2026-10-09T09:27:47Z", "reviews": [ { "id": "2610.13276", "title": "CopRoute: Dependence-Aware Conformal Routing for Language Models", "published": "2026-10-09T09:27:47Z", "decision": "reviewed_deferred", "reason": "designcorrectnessmatrix用greedytally按新增覆盖排序，frozen query stopping score形成nestedmodelsets；独立split calibration选最小certifiedcutoff，否则NON-CERTIFIED。; RouterBench/RouterEval matchedall-missrisk；GSM8K单uniqueprompt、MBPPfullpoolmiss过高被排除；directtopmodel有更低cost但不能certify。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13276v1#S3", "https://arxiv.org/html/2610.13276v1#S4", "https://arxiv.org/html/2610.13276v1#S5", "https://arxiv.org/html/2610.13276v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13276v1", "sha256": "c6595f73df717e2e6451d6cfbb03677bd208211323e0f77044200425fa1c19b5" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13273", "title": "Agent4Spec: An LLM-based Agentic System to Optimize API Specifications through Automated Validation and Market Analysis", "published": "2026-10-09T09:25:30Z", "reviews": [ { "id": "2610.13273", "title": "Agent4Spec: An LLM-based Agentic System to Optimize API Specifications through Automated Validation and Market Analysis", "published": "2026-10-09T09:25:30Z", "decision": "reviewed_deferred", "reason": "已读摘要、Agent4Spec与实验：orchestrator二选一syntax validation或市场缺口research后revision，curated300公共+200工业API规格。syntax与修订adoption不等价真实endpoint可执行成功或生产A/B，下一批workflow infra。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "IV Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "III Agent4Spec", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13273v1", "sha256": "32c3885318a2960e6841c9fd269e86cdb682e4bd8411822fa8de8a45665f72d0", "local_html_sha256": "32c3885318a2960e6841c9fd269e86cdb682e4bd8411822fa8de8a45665f72d0", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13273v1", "result": "retrieved", "sha256": "32c3885318a2960e6841c9fd269e86cdb682e4bd8411822fa8de8a45665f72d0" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13272", "title": "OWLET: On-Policy Weighted Learning via Evidence Transfer for Video Large Language Model Post-Training", "published": "2026-10-09T09:23:21Z", "reviews": [ { "id": "2610.13272", "title": "OWLET: On-Policy Weighted Learning via Evidence Transfer for Video Large Language Model Post-Training", "published": "2026-10-09T09:23:21Z", "decision": "reviewed_deferred", "reason": "OWLET student用uniformframes自己on-policy rollout，same-size teacher对同response用global+questionlocalprivilegedview评分；tokenlogprobcontrast只重分positiveadvantage，nonpositive不变；verifieraccepted响应cosinehiddenstate alignment。; Video-Evidence96K训练、四Qwen3.5scale五video benchmarks，预算/teacher/ECR/SSA ablations及其他Qwen3VL外推。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13272v1#S3", "https://arxiv.org/html/2610.13272v1#S4", "https://arxiv.org/html/2610.13272v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13272v1", "sha256": "5f6f11e5cd2c4e3a22b7c5cc994d1affb5086049d5f8089599b49c95965565f8" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13267", "title": "Forgetting is Not Always Bad: A Neuro-Inspired Memory Repair Mechanism for Poisoned LLM Agents", "published": "2026-10-09T09:17:00Z", "reviews": [ { "id": "2610.13267", "title": "Forgetting is Not Always Bad: A Neuro-Inspired Memory Repair Mechanism for Poisoned LLM Agents", "published": "2026-10-09T09:17:00Z", "decision": "reviewed_deferred", "reason": "已读DREAM方法/实验片段：embedding Gaussian扰动重激活、多维异常诊断、选择性记忆抑制与良性保留，EHR/ReAct/多agent poisoning benchmarks。需实际embedding/拓扑实现并分离安全效用，不能以固定恶意标签fixture代替诊断；下一批。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Methods", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments and Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13267v1", "sha256": "8f3527562fb199a9e3df240782702c320681dcbcf4d97cb759a87e53566a8f66", "local_html_sha256": "8f3527562fb199a9e3df240782702c320681dcbcf4d97cb759a87e53566a8f66", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13267v1", "result": "retrieved", "sha256": "8f3527562fb199a9e3df240782702c320681dcbcf4d97cb759a87e53566a8f66" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13263", "title": "SheafPrune: Sheaf-Guided Graph Propagation for Visual Token Pruning", "published": "2026-10-09T09:15:08Z", "reviews": [ { "id": "2610.13263", "title": "SheafPrune: Sheaf-Guided Graph Propagation for Visual Token Pruning", "published": "2026-10-09T09:15:08Z", "decision": "reviewed_deferred", "reason": "SheafPrune modalitymaps+sparse textvisualgraph传relevance，独立visualgraph propagate并node-degree纠正，ratio-conditioned4.5Mselector经frozenVLM语言loss训练，predecoder一次hardtopk，structuralnewline不占budget。; 四backbone十benchmark七ratio unifiedbaselines；selector之后prefill缩短，visualencoding/projection已发生；depth增加并非普遍更好。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13263v1#S3", "https://arxiv.org/html/2610.13263v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13263v1", "sha256": "e171584ffbb0be90318415acaa2d35af54d5534e81046e8b57a8046a453dfded" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13262", "title": "Dual-Stream Memory for Vision-Language-Action Policies", "published": "2026-10-09T09:14:59Z", "reviews": [ { "id": "2610.13262", "title": "Dual-Stream Memory for Vision-Language-Action Policies", "published": "2026-10-09T09:14:59Z", "decision": "reviewed_deferred", "reason": "taskagnosticGatedDeltaNet把history成1temporal token，progress+actionconditionedfeature reconstruction监督；taskconditionretrieverfixedthreshold选真实frames，linear multiscalefusion16spatialtokens；冻结compressor后jointpolicyretrieval训。; 16RoboMME/4realrobot，50episodes/task×三evalseed；cup-swap Shell比baseline差，identitytracking未解决。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13262v1#S3", "https://arxiv.org/html/2610.13262v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13262v1", "sha256": "e6267822f2ea9cc24ac0975342d2ff4c588109b90ca4f709f37e8998af441ff9" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13261", "title": "Data-Adaptive Mahalanobis Metric Learning for Cross-Head Attention in Transformers", "published": "2026-10-09T09:11:08Z", "reviews": [ { "id": "2610.13261", "title": "Data-Adaptive Mahalanobis Metric Learning for Cross-Head Attention in Transformers", "published": "2026-10-09T09:11:08Z", "decision": "reviewed_deferred", "reason": "DAMCHA input-dependentfullcross-headMahalanobis-like矩阵保留offdiagonalhead交互，stack-wise参数共享；数学严格扩展仅pre-softmaxscores在generator表达力/curvature假设下。; 小decoder CIFARpixelAR/WMT14/CommonGen与Enwik8/Text8/DCLM-Mix LM100Ksteps，heldoutvalidationPPL，不是frontier规模生产serving。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13261v1#S3", "https://arxiv.org/html/2610.13261v1#S4", "https://arxiv.org/html/2610.13261v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13261v1", "sha256": "0758e23c79425613df5f7da45253209d5037fac5f570cef0743fac36cc4d71be" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13259", "title": "Beyond a Single Crop: Contextualized Embeddings for Player Re-Identification", "published": "2026-10-09T09:07:45Z", "reviews": [ { "id": "2610.13259", "title": "Beyond a Single Crop: Contextualized Embeddings for Player Re-Identification", "published": "2026-10-09T09:07:45Z", "decision": "reviewed_deferred", "reason": "ContextReID用同frame co-occurring crops+precedingframes视觉context生成单playerembedding；contextuallearning/retrievalloss，推理只detections/frame membership无personID/tracklets。; SportsMOT/DanceTrack/SoccerNet GSR三MOT转cropretrieval Rank1/mAP；t-SNE仅qualitativecohesion。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13259v1#Sx3", "https://arxiv.org/html/2610.13259v1#Sx4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13259v1", "sha256": "a937d379d3641f36fa389d7b72da20717c19fe6e4f929468409d9a11d5b66ddf" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13257", "title": "CAST: Credit Alignment for Cross-Tokenizer On-Policy Distillation", "published": "2026-10-09T09:03:37Z", "reviews": [ { "id": "2610.13257", "title": "CAST: Credit Alignment for Cross-Tokenizer On-Policy Distillation", "published": "2026-10-09T09:03:37Z", "decision": "accepted", "reason": "跨 tokenizer 文本对齐后的 credit 与未对齐 KL 是明确核心机制，主节方法和控制实验已审，纳入用户冻结批次；保留原论文评测选择偏差，不移植其 test-selection 做法。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "official abstract; full PDF pages 4-8 and page 18 relevant method/setup/results/selection protocol; not every appendix or qualitative example", "review_artifact": "research-audits/oct12-review-closure/frozen-post-training-reviews.json", "evidence_locations": [ { "kind": "method", "heading": "3.1 Preliminaries", "pdf_pages": [ 4 ], "figure": 3, "scope": "teacher retokenization of the actual student-decoded text, alignment chunks, projected scores and separate six-token credit chunks" }, { "kind": "method", "heading": "3.2 Uncertainty-Smoothness Credit Assignment", "pdf_pages": [ 5 ], "equations": [ 1, 2, 3, 4, 5, 6, 7 ], "scope": "full page; entropy times inverse local teacher-signal variation, stop-gradient weights and conserved chunk advantage" }, { "kind": "method", "heading": "3.3 Selective Token Anchoring and Optimization", "pdf_pages": [ 6 ], "equations": [ 8, 9, 10, 11 ], "scope": "unaligned-token initial-policy KL, aligned PPO objective and independent multiple rollouts" }, { "kind": "evaluation", "heading": "4.1 Experimental Setup; 4.2 Main Results; 4.3 Discussions", "pdf_pages": [ 6, 7, 8 ], "tables": [ 1, 2, 3, 4, 5, 6 ], "scope": "heterogeneous/shared tokenizer settings, matched uniform/shuffled credit, anchoring and rollout controls; three evaluated seed summaries" }, { "kind": "evaluation_integrity", "heading": "Appendix C: Checkpoint selection; Evaluation protocol and run variability", "pdf_pages": [ 18 ], "table": 9, "characters_read": [ 0, 2983 ], "scope": "all six benchmarks used to choose highest-Avg-6 checkpoints/seed-checkpoint pairs; original paper explicitly warns optimistic selection" } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/pdf/2610.13257v1", "sha256": "a930beee09025e2abb6a7c82ca412c2cdbdc7fd9aba16eedfcdee19af7eb6e85", "local_path": "tmp/pdfs/2610.13257.pdf", "local_pdf_sha256_matches": true, "pdf_pages": 22, "version_evidence": "PDF page 1 identifies arXiv:2610.13257v1 and 9 Oct 2026" }, "new_implementation_inferred_from_review": false } ], "discovery_origin": "independent-targeted-supplement", "implementation_status": "registered_scoped_runtime; actual acceptance tracked separately", "status": "accepted", "implementation_key": "cast-opd", "accepted_scope": "bounded faithful mechanism/runtime or evaluation infrastructure; not original capability replication" }, { "id": "2610.13253", "title": "Tabular Foundation Models Are Effectively Shallow", "published": "2026-10-09T09:01:42Z", "reviews": [ { "id": "2610.13253", "title": "Tabular Foundation Models Are Effectively Shallow", "published": "2026-10-09T09:01:42Z", "decision": "reviewed_deferred", "reason": "用unlabeled calibration internalHs→He ridgeclosedform affine translator替换contiguousblockwindow，而非identityskip；teacheroutputKL在calibration选择最大压缩窗口。; 8TFM/171classificationdataset、临床/回归补充；RTX5090 GPU/CPU效率，oneblockregressors通常不保ranking。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13253v1#S3", "https://arxiv.org/html/2610.13253v1#S4", "https://arxiv.org/html/2610.13253v1#S5", "https://arxiv.org/html/2610.13253v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13253v1", "sha256": "195239740656e8b04555e9a18615f9a16620418630700c6e1f177bc68ac017c0" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13252", "title": "Vision Transformers Learn Word Meaning from Images Alone", "published": "2026-10-09T09:00:08Z", "reviews": [ { "id": "2610.13252", "title": "Vision Transformers Learn Word Meaning from Images Alone", "published": "2026-10-09T09:00:08Z", "decision": "reviewed_deferred", "reason": "真实visionencoder对renderedwordimages减matchedcontrol成worddirection，控制拼写/字体并比较nearestsemanticneighbors、countrycapitaloffset和visualclasscosine；另text/speechcontrastivefinetune。; 14models，多大DINOv3有结构而其他largeSSL不呈现；DINOH比7Bwordclass对齐强原因未解释。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13252v1#S3", "https://arxiv.org/html/2610.13252v1#S4", "https://arxiv.org/html/2610.13252v1#S5", "https://arxiv.org/html/2610.13252v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13252v1", "sha256": "ba2777699f47685a586ecc144eedf0655f596465a44f0eb7583963f3b60ad0ff" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13239", "title": "Jag: Direct Box Prediction for Efficient Visual Grounding", "published": "2026-10-09T08:54:42Z", "reviews": [ { "id": "2610.13239", "title": "Jag: Direct Box Prediction for Efficient Visual Grounding", "published": "2026-10-09T08:54:42Z", "decision": "reviewed_deferred", "reason": "Jag真实Qwen3.5-0.8B image+refexpression末个有效inputtoken hidden读轻regressionhead同预测4continuouscoords；headwarmup后geometryloss联合adapt语言backbone/visualmerger/head，无coordtokendecode。; RefCOCO/+ /g officialsplit combinedtraining、AccIoU.5；base完整适配对比和head-only消融，含wrong-instance/extent失败。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13239v1#S3", "https://arxiv.org/html/2610.13239v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13239v1", "sha256": "1ea9bb4e2146a06c219f15cdb7d215993b064ee1cb753fe9992258befbbb283a" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13227", "title": "Understanding and Mitigating Cross-Session Attacks via Persistent Artifacts in Agent Skill Composition", "published": "2026-10-09T08:45:27Z", "reviews": [ { "id": "2610.13227", "title": "Understanding and Mitigating Cross-Session Attacks via Persistent Artifacts in Agent Skill Composition", "published": "2026-10-09T08:45:27Z", "decision": "reviewed_deferred", "reason": "已读SkillRelay threat model与评测：跨session持久artifact的IP/DBV/EP，270case30场景，guard恢复provenance并pre-execution advisory，但终权限仍在agent。安全task completion与ASR双指标，下一批隔离评测不无权执行第三方skill。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3. Threat Model and Attack Patterns", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5. Empirical Evaluation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13227v1", "sha256": "2fe7422687b51c0c60781acd33ef55a7b6bc8f98c26fd6217d1cb0cc9ba6980b", "local_html_sha256": "2fe7422687b51c0c60781acd33ef55a7b6bc8f98c26fd6217d1cb0cc9ba6980b", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13227v1", "result": "retrieved", "sha256": "2fe7422687b51c0c60781acd33ef55a7b6bc8f98c26fd6217d1cb0cc9ba6980b" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13208", "title": "How Much Computation Does Every Forecast Need? Structured Block Skipping in Two Time-Series Foundation Models", "published": "2026-10-09T08:26:14Z", "reviews": [ { "id": "2610.13208", "title": "How Much Computation Does Every Forecast Need? Structured Block Skipping in Two Time-Series Foundation Models", "published": "2026-10-09T08:26:14Z", "decision": "reviewed_deferred", "reason": "真实ChronosBolt/TImesFM2.5原生block早返回bypass，allblocks-enable exactbaseline；staticmasks只calibration选，executioncounters与syncGPU测成本。; 四series1088heldoutwindows A100，moderateaggregate近中性掩盖Traffic/Weather异质；equalbudgetperwindoworacle retrospectivelybetter但LODO routing失败。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13208v1#S3", "https://arxiv.org/html/2610.13208v1#S4", "https://arxiv.org/html/2610.13208v1#S5", "https://arxiv.org/html/2610.13208v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13208v1", "sha256": "cb10c5248ff9601d6401cf949704418cb329ffc345298eb4df441e8c941de2e9" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13202", "title": "S$^3$R: Fast and Faithful RL Rollouts via Self-Sparse Speculative Decoding", "published": "2026-10-09T08:22:40Z", "reviews": [ { "id": "2610.13202", "title": "S$^3$R: Fast and Faithful RL Rollouts via Self-Sparse Speculative Decoding", "published": "2026-10-09T08:22:40Z", "decision": "reviewed_deferred", "reason": "S3R同参数sparse drafting+dense exact speculative verification，保留完整dense group分布及KV-aware scheduler；8H100 DAPO math GRPO throughput对照，下一批需实际dense correction和GPU latency/分布验证。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Problem Formulation and Design Principles", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "6 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13202v1", "sha256": "2c4704fec367030df4d39d71da112f0f5dd160df8bf803e558a25adf417c33e1", "local_html_sha256": "2c4704fec367030df4d39d71da112f0f5dd160df8bf803e558a25adf417c33e1", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13202v1", "result": "retrieved", "sha256": "2c4704fec367030df4d39d71da112f0f5dd160df8bf803e558a25adf417c33e1" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.13194", "title": "Janus: Enhancing Long-Horizon Autonomous Research with Co-Evolving World Models", "published": "2026-10-09T08:13:28Z", "reviews": [ { "id": "2610.13194", "title": "Janus: Enhancing Long-Horizon Autonomous Research with Co-Evolving World Models", "published": "2026-10-09T08:13:28Z", "decision": "accepted", "reason": "用户点名冻结16范围；selected defining-method/experiment原文证据核验，具体实现/运行等级独立记录。; 研究图中稀疏soft Bradley–Terry tournament预测VI/IG/EV再refine、真实trial，反馈驱动intra/cross/prediction/exploration/coordination技能演化。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "实施期间读取摘要、中心方法/训练/评测及部分相关附录；本次复用这些已读来源并重新校验官方HTML字节SHA和v1日期，不声明全文每段已读。", "review_artifact": "research-audits/oct12-review-closure/frozen-agent-reviews.json", "evidence_locations": [ "3.1 Problem Formulation", "3.2 The Predict-Verify-Evolve Loop", "3.3 World Model Guided Exploration Strategy", "3.4 Experience-Driven Co-Evolution", "Appendix A Formal Definitions of Research Priority Criteria", "4.1 Experimental Setup", "4.2 Performance Analysis", "4.3 Ablation Study", "4.4 Framework Analysis", "Appendix D Metrics", "Appendix E Implementation Details" ], "dateproof": { "url": "https://arxiv.org/abs/2610.13194", "sha256": "e08fb761ff05a7b820c7abc08b6f2fa17ffa55b2d66ee25507892bfc26d91902", "location": "Submission history / [v1]", "first_publication_utc": "2026-10-09T08:13:28Z" }, "source_proof": { "url": "https://arxiv.org/html/2610.13194v1", "sha256": "a6e2f23883bb8f68a01d002000b2f628e90829719f0c229f591d0c7dfd9d57d9", "byte_hash_verified_on": "2026-10-12" }, "new_implementation_inferred_from_review": false } ], "discovery_origin": "independent-targeted-supplement", "implementation_status": "registered_scoped_runtime; actual acceptance tracked separately", "status": "accepted", "implementation_key": "janus", "accepted_scope": "bounded faithful mechanism/runtime or evaluation infrastructure; not original capability replication" }, { "id": "2610.13172", "title": "RoleWorld: Knowledge Transfer in World Modeling with Functional Role Abstractions", "published": "2026-10-09T07:58:45Z", "reviews": [ { "id": "2610.13172", "title": "RoleWorld: Knowledge Transfer in World Modeling with Functional Role Abstractions", "published": "2026-10-09T07:58:45Z", "decision": "reviewed_deferred", "reason": "已读RoleWorld方法/实验：teacher从真实trajectory抽role prototypes/executable rules，target actor不更新权重、固定rules而在线推断role与参数，MARS/Crafter22achievement。下一批需真实环境反馈与未知behavior transfer，非固定world truth。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiment", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13172v1", "sha256": "5c2eac8b5e01075c005f0653d698075bf8f4430574d07588bcaaf86a8a8e5e5b", "local_html_sha256": "5c2eac8b5e01075c005f0653d698075bf8f4430574d07588bcaaf86a8a8e5e5b", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13172v1", "result": "retrieved", "sha256": "5c2eac8b5e01075c005f0653d698075bf8f4430574d07588bcaaf86a8a8e5e5b" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13166", "title": "Scaling Agentic Data for Long-Horizon Terminal Intelligence", "published": "2026-10-09T07:55:55Z", "reviews": [ { "id": "2610.13166", "title": "Scaling Agentic Data for Long-Horizon Terminal Intelligence", "published": "2026-10-09T07:55:55Z", "decision": "reviewed_deferred", "reason": "已读TerminalHorizon方法/实验：独立execution发现合成环境预先完成的关键工作，再reconstruct并扩展同goal任务；1.5k环境、Qwen3.5-27B SFT与12bench。production settings是任务评测不证明线上A/B；下一批须隔离environment gold verifier与agent可见包。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13166v1", "sha256": "b5870d1910a879149f8a236f81cf989117ab24ee43edbf138bd275198bae5b86", "local_html_sha256": "b5870d1910a879149f8a236f81cf989117ab24ee43edbf138bd275198bae5b86", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13166v1", "result": "retrieved", "sha256": "b5870d1910a879149f8a236f81cf989117ab24ee43edbf138bd275198bae5b86" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13154", "title": "From Glance to Focus: Mitigating Recognition Laziness in MLLMs for Fine-Grained Recognition", "published": "2026-10-09T07:42:32Z", "reviews": [ { "id": "2610.13154", "title": "From Glance to Focus: Mitigating Recognition Laziness in MLLMs for Fine-Grained Recognition", "published": "2026-10-09T07:42:32Z", "decision": "reviewed_deferred", "reason": "LIVE trainingfree以query视觉relevance+intercategoryaffinitygraph做adaptivecandidate inference，VLMselfreflection构建categoryspecificheaders/细粒度comparativeskills，再按uncertainty/confusion选择detailguidance。; 13 recognitionbench多backbone；fixedtopK部分Flower设置更差，compactcandidate必须兼顾coverage。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13154v1#S3", "https://arxiv.org/html/2610.13154v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13154v1", "sha256": "57c147fd6e17670b2c36f1b6e3fbfc20841b2417ad46c9af82ee9c997ba896f8" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13152", "title": "Quality, Not Just Outcome: A Structural Scoring Framework for Selecting Coding-Agent Trajectories for Supervised Fine-Tuning", "published": "2026-10-09T07:40:42Z", "reviews": [ { "id": "2610.13152", "title": "Quality, Not Just Outcome: A Structural Scoring Framework for Selecting Coding-Agent Trajectories for Supervised Fine-Tuning", "published": "2026-10-09T07:40:42Z", "decision": "reviewed_deferred", "reason": "已读TQS结构分数/实验片段：线性时间trajectory信息focus、intent-action和policy dynamics，无LLM judge；155144成功trajectory排序后实际SFT对比random/bottom。下一批需真实训练与query/repo分离，结构分数不能单独证明能力。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Trajectory Quality Score", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13152v1", "sha256": "bd23a4a4ec2216036f804ed534be5047281d02bd02b588cf21a619b0a78f98ea", "local_html_sha256": "bd23a4a4ec2216036f804ed534be5047281d02bd02b588cf21a619b0a78f98ea", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13152v1", "result": "retrieved", "sha256": "bd23a4a4ec2216036f804ed534be5047281d02bd02b588cf21a619b0a78f98ea" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13144", "title": "Competence-Aware Orchestration for LLM-Based Multi-Agent Systems", "published": "2026-10-09T07:33:42Z", "reviews": [ { "id": "2610.13144", "title": "Competence-Aware Orchestration for LLM-Based Multi-Agent Systems", "published": "2026-10-09T07:33:42Z", "decision": "reviewed_deferred", "reason": "已读competence orchestration方法/实验：固定probe capability fingerprint加query-conditioned模型，learned role activation与communication graph；6 reasoning/codebench及unseen backbones。下一批须训练真实competence predictor，probe/test隔离而非gold推断能力。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13144v1", "sha256": "a48156db716000bd3fc9648721bd9e39d8805db68ae33ad5731521141d126645", "local_html_sha256": "a48156db716000bd3fc9648721bd9e39d8805db68ae33ad5731521141d126645", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13144v1", "result": "retrieved", "sha256": "a48156db716000bd3fc9648721bd9e39d8805db68ae33ad5731521141d126645" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13143", "title": "Measurable Semantic 3D Gaussian Splatting with Joint Region and Depth Supervision for Interactive Building Inspection", "published": "2026-10-09T07:33:37Z", "reviews": [ { "id": "2610.13143", "title": "Measurable Semantic 3D Gaussian Splatting with Joint Region and Depth Supervision for Interactive Building Inspection", "published": "2026-10-09T07:33:37Z", "decision": "rejected_or_out_of_scope", "reason": "MVS reference直接监督Gaussian center pointdepth而非opacity混合renderdepth，foregroundmask限制RGB/depthdomain；semanticfeature绑Gaussianprimitive，AEC SplatTwin确定性measurement平台由MLLM解释结果。; QuakeCity synthetic和HKUPB smartphone；depthMAE对同MVS训练reference是几何一致性；独立window scale校准后7targets外部manualmeasurement。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "PDF pages1..6 first2300chars each; pages14,15,16,25,27,43,46 first1100chars each; not complete59pages", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/pdf/2610.13143v1#page=14", "https://arxiv.org/pdf/2610.13143v1#page=25", "https://arxiv.org/pdf/2610.13143v1#page=43" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/pdf/2610.13143v1", "sha256": "9c45adc270721db3badb54c823e18dc2f0a6f11d2a6d6167f454c6a253b118c0" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.13128", "title": "Locate, Judge, Read: Role-Specialized Tool Agents over a Knowledge-Point Index for Open-Domain Multimodal Document QA", "published": "2026-10-09T07:21:36Z", "reviews": [ { "id": "2610.13128", "title": "Locate, Judge, Read: Role-Specialized Tool Agents over a Knowledge-Point Index for Open-Domain Multimodal Document QA", "published": "2026-10-09T07:21:36Z", "decision": "reviewed_deferred", "reason": "KnowForge indexatomicpoints/hybrid retrieval，Locator只search，codebridgejudge强制二hop、typedboundedpool合并rank，DeepReader只读真实text/image/table，synthesizer+formatcheck。; M3DocVQA/MMLongBenchDoc/MultiHopRAG fullruns和mid300sweeps；多数数字singlefullrun，视觉reader在textcorpus贡献0；selfhostserver共享productiontraffic仅说明latency环境。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13128v1#S3", "https://arxiv.org/html/2610.13128v1#S4", "https://arxiv.org/html/2610.13128v1#S5", "https://arxiv.org/html/2610.13128v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13128v1", "sha256": "6e29c15ecace7d0caf92a86f5dff11c1829b724ec2c5b9796de0538fd4b9a04c" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13116", "title": "DRYLAB-Bench: Benchmarking the Biosecurity Risk of Agentic LLM Mutation Design", "published": "2026-10-09T07:09:16Z", "reviews": [ { "id": "2610.13116", "title": "DRYLAB-Bench: Benchmarking the Biosecurity Risk of Agentic LLM Mutation Design", "published": "2026-10-09T07:09:16Z", "decision": "rejected_or_out_of_scope", "reason": "已读摘要、构建与evaluation高层片段：是高风险生物mutation设计能力测量，不属于当前通用agent实现批；涉及增强病原性/耐药风险，不运行设计优化或工具链，保留安全风险研究引用。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Evaluation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 DRYLAB-Bench Composition and Construction", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13116v1", "sha256": "de7f4bd26a0cb2f9a82abc6dda745e089e51e5d07bc3a640b416d34e55824008", "local_html_sha256": "de7f4bd26a0cb2f9a82abc6dda745e089e51e5d07bc3a640b416d34e55824008", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13116v1", "result": "retrieved", "sha256": "de7f4bd26a0cb2f9a82abc6dda745e089e51e5d07bc3a640b416d34e55824008" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.13111", "title": "What Do Visual-Head Discovery Scores Actually Find? Attention Source, Necessity, and Random-Control Pitfalls in Vision-Language Models", "published": "2026-10-09T07:06:17Z", "reviews": [ { "id": "2610.13111", "title": "What Do Visual-Head Discovery Scores Actually Find? Attention Source, Necessity, and Random-Control Pitfalls in Vision-Language Models", "published": "2026-10-09T07:06:17Z", "decision": "reviewed_deferred", "reason": "对knowntargetregion attentionmass排名，querytype/corpus/attention-source 2×3×2 discovery；steeringCERS与masknecessityNCS分别对layer及attentionmassmatchedrandomheadsets验证。; Qwen3VL8B 12sets、250random核心对照与其他VLMreplication；steerability无特异优势，necessity按attention source明显不同。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13111v1#S3", "https://arxiv.org/html/2610.13111v1#S4", "https://arxiv.org/html/2610.13111v1#S5", "https://arxiv.org/html/2610.13111v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13111v1", "sha256": "2c19e4920ab61aedcde569e7b7bd5eaf8fefd977a977a331ef6e0c3f29ee3bf8" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13106", "title": "Scaling RNA Foundation Models through Iterative Refinement", "published": "2026-10-09T06:59:15Z", "reviews": [ { "id": "2610.13106", "title": "Scaling RNA Foundation Models through Iterative Refinement", "published": "2026-10-09T06:59:15Z", "decision": "reviewed_deferred", "reason": "EcoRNA prelude→sharedinputconditionTransformercore迭代→coda，训练跨depth maskedreconstruction/state reg/one-sideddeep-loss以shallowreference锚定observednucleotide；参数不随K增长。; 80M RNAGym70assays、sequencefamily/structure/chemicalprobing；depthsweep固定checkpoint/head，coarselabel较早plateau而denseoutputs后续小增益。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13106v1#S2", "https://arxiv.org/html/2610.13106v1#S3", "https://arxiv.org/html/2610.13106v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13106v1", "sha256": "2674df0b300e6f24153bd8d9c39677a4a8b86a35dd7a823e8e1c534c63d6fa20" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13102", "title": "From 2.1M Hours of Raw Broadcast Audio to Finnish Speech SSL: Confidence Filtering and Architectural Analysis at Scale", "published": "2026-10-09T06:57:05Z", "reviews": [ { "id": "2610.13102", "title": "From 2.1M Hours of Raw Broadcast Audio to Finnish Speech SSL: Confidence Filtering and Architectural Analysis at Scale", "published": "2026-10-09T06:57:05Z", "decision": "reviewed_deferred", "reason": "Finnishwav2vec2 SSL用ASRconfidence筛pretrain音频而非仅VAD/LID存在判断，backendsTransformer/Macaron/Conformer/EBranchformer按参数与pretrain数据matched对照。; 2.1M原broadcasthours六ASRtestsets、95/317/965M模型，512MI250Xpretrain；部分Xlarge domain比Large退化，conv优势随容量减弱。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13102v1#S3", "https://arxiv.org/html/2610.13102v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13102v1", "sha256": "70b1fe29163bf100c760e31e736b65de0d58120f0b092b99f7923258b36243fc" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13092", "title": "When Faster Decisions Do Not Make Better Agents: An Empirical Study of Jev in Agent Workflows", "published": "2026-10-09T06:45:23Z", "reviews": [ { "id": "2610.13092", "title": "When Faster Decisions Do Not Make Better Agents: An Empirical Study of Jev in Agent Workflows", "published": "2026-10-09T06:45:23Z", "decision": "accepted", "reason": "冻结16范围已有Jev workflow evaluation infrastructure：配对交付结果、decision/task latency、含失败的cost与恢复审计，CPU非Jev模型。已读原文评测且找到作者anonymous链接，未验证可运行源码；原文aggregate数值不冒充本地run。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Empirical Study Design", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 6193 ] }, { "heading": "5 Results and Analysis", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 13351 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13092v1", "sha256": "3db7474953cbe5a3052fb22793038f2383bdd0e7c775af06494a5e8501646d18", "local_html_sha256": "3db7474953cbe5a3052fb22793038f2383bdd0e7c775af06494a5e8501646d18", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13092v1", "result": "retrieved", "sha256": "3db7474953cbe5a3052fb22793038f2383bdd0e7c775af06494a5e8501646d18" } ] }, "new_implementation_inferred_from_review": false }, { "id": "2610.13092", "title": "When Faster Decisions Do Not Make Better Agents: An Empirical Study of Jev in Agent Workflows", "published": "2026-10-09T06:45:23Z", "decision": "accepted", "reason": "用户点名冻结16范围；selected defining-method/experiment原文证据核验，具体实现/运行等级独立记录。; 比较agent交付任务结果及decision/task latency、成本/成功交付、失败和恢复；不是新model或agent算法。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "实施期间读取摘要、中心方法/训练/评测及部分相关附录；本次复用这些已读来源并重新校验官方HTML字节SHA和v1日期，不声明全文每段已读。", "review_artifact": "research-audits/oct12-review-closure/frozen-agent-reviews.json", "evidence_locations": [ "4.1 Research Questions", "4.2 Experimental Setup", "4.3 Evaluation and Metrics", "Appendix A Evaluation and Statistical Methods", "5.1 Effectiveness", "5.2 Efficiency Transfer", "5.3 Failure and Recovery", "Limitations", "Appendix C Trace Analysis and Case Evidence" ], "dateproof": { "url": "https://arxiv.org/abs/2610.13092", "sha256": "174dcb8eb2d37831d8a22ea3f6746997c340e74bb373100f3926c0d0b706daa1", "location": "Submission history / [v1]", "first_publication_utc": "2026-10-09T06:45:23Z" }, "source_proof": { "url": "https://arxiv.org/html/2610.13092v1", "sha256": "3db7474953cbe5a3052fb22793038f2383bdd0e7c775af06494a5e8501646d18", "byte_hash_verified_on": "2026-10-12" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "accepted", "implementation_key": "jev-workflow", "accepted_scope": "bounded faithful mechanism/runtime or evaluation infrastructure; not original capability replication" }, { "id": "2610.13071", "title": "CAPE - Context-Aware Prioritization Engine for Vulnerability Scoring and Management", "published": "2026-10-09T06:30:34Z", "reviews": [ { "id": "2610.13071", "title": "CAPE - Context-Aware Prioritization Engine for Vulnerability Scoring and Management", "published": "2026-10-09T06:30:34Z", "decision": "rejected_or_out_of_scope", "reason": "已读CAPE打分与评测：依据deployment reachability/centrality/exploitability排序依赖CVEs，30agentic应用只是评测软件对象。中心算法不是LLM agent或模型训练，artifact承诺以后发布，当前范围外安全工程。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "V Priority Scoring Model", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "VI Evaluation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13071v1", "sha256": "9064488cc46f1a478667f1bec814c8602f719f1463a7393844dae70951780fa2", "local_html_sha256": "9064488cc46f1a478667f1bec814c8602f719f1463a7393844dae70951780fa2", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13071v1", "result": "retrieved", "sha256": "9064488cc46f1a478667f1bec814c8602f719f1463a7393844dae70951780fa2" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.13066", "title": "Evolving Creatives with Clicks: CTR-Driven Advertising image Editing via Attribute Disentanglement and Preference Alignment", "published": "2026-10-09T06:26:24Z", "reviews": [ { "id": "2610.13066", "title": "Evolving Creatives with Clicks: CTR-Driven Advertising image Editing via Attribute Disentanglement and Preference Alignment", "published": "2026-10-09T06:26:24Z", "decision": "reviewed_deferred", "reason": "不是因无AB词拒绝：原文已有真实线上CTR对照。冻结16之外，且私有click训练数据/作者代码尚不可核验，明确留待下一批资源条件满足再实施。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "官方摘要、Methods三小节、Datasets、实验设置、线上协议及Tables1/2/4/5/6与消融；未宣称逐段读附录或全部全文。", "review_artifact": "research-audits/oct12-review-closure/ecc-review.json", "evidence_locations": [ "Methods: CTR-Aware Visual Attribute Mining, equations (1)", "Methods: Instruction-Guided Editing Learning, equation (2)", "Methods: Preference Optimization, equations (3)-(5)", "Datasets: SFT/preference construction and privileged-reference exclusion", "Experiments: Online Evaluation Protocol and Evaluation Metrics, equation (6)", "Online Results: Tables 1-2", "Order Robustness Analysis: Table 4", "Validation of CTR-Sensitive Visual Attributes: Table 5", "Ablation Study: Table 6" ], "dateproof": { "url": "https://arxiv.org/abs/2610.13066", "sha256": "acabf6fa82cdac7a603228a0d9f9bd768ef79d0cb6fa1f17818fc264944ec9a9", "location": "Submission history / [v1]" }, "source_proof": { "url": "https://arxiv.org/html/2610.13066v1", "sha256": "355fa2a4dbb4aca5a058339af9f4f18188616128d47975b166031e23e6fe8be5", "byte_hash_verified": true }, "new_implementation_inferred_from_review": false } ], "status": "reviewed_deferred", "discovery_origin": "independent-targeted-supplement", "implementation_status": "not_implemented_in_frozen16", "accepted_scope": "conditional candidate independently reviewed; asset availability deferred, no silent omission" }, { "id": "2610.13065", "title": "TestSuiteBench: Coding Agents Struggle to Generate Strong Test Suites from Specifications", "published": "2026-10-09T06:26:02Z", "reviews": [ { "id": "2610.13065", "title": "TestSuiteBench: Coding Agents Struggle to Generate Strong Test Suites from Specifications", "published": "2026-10-09T06:26:02Z", "decision": "reviewed_deferred", "reason": "已读TestSuiteBench构建与实验：165 app、逐requirement高选择性mutants，测试validity/reference pass/mutation score分开；agent生成tests时不得读取reference实现或test。适合下一批评测infra，低于60% mutation不能被88% reference pass掩盖。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Experimental Evaluation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 Constructing TestSuiteBench", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13065v1", "sha256": "18fb037c9c1edf3bab8cef7e2e757dbdd44751560c695bb37745d4e882626687", "local_html_sha256": "18fb037c9c1edf3bab8cef7e2e757dbdd44751560c695bb37745d4e882626687", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13065v1", "result": "retrieved", "sha256": "18fb037c9c1edf3bab8cef7e2e757dbdd44751560c695bb37745d4e882626687" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13064", "title": "CB-Orchestrator: Adaptive Workflow Optimization for LLM Agents via Contextual Bandits", "published": "2026-10-09T06:25:19Z", "reviews": [ { "id": "2610.13064", "title": "CB-Orchestrator: Adaptive Workflow Optimization for LLM Agents via Contextual Bandits", "published": "2026-10-09T06:25:19Z", "decision": "reviewed_deferred", "reason": "候选workflow为可执行asyncPythonclass/operator组合，evolutionarysearch建pool，再contextualbandit探索执行workflow收reward训练selector，infer无explorationbonus并Wilsonfilter未充分trialworkflow。; GSM8K/MATH/HumanEval/MBPP/GAIA，training-token开销同时报告是否包括poolconstruction。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13064v1#S3", "https://arxiv.org/html/2610.13064v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13064v1", "sha256": "47627ed17014e8d3a9883c79c595b555a8d01db42e07f84ea723fa5c8cd7ad8d" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13060", "title": "KV-Repurpose: A Testbed for Evaluating Non-Prefix KV-Cache Reuse", "published": "2026-10-09T06:17:10Z", "reviews": [ { "id": "2610.13060", "title": "KV-Repurpose: A Testbed for Evaluating Non-Prefix KV-Cache Reuse", "published": "2026-10-09T06:17:10Z", "decision": "reviewed_deferred", "reason": "vLLM扩展server+cacheproxy做非prefixsegmentKV复用，positional correction、segmentschedule/promptcapacity/LRU/NVMe监控；pseudoRAGquestioncontextdatabase形成workload。; GPTOSS120B单H200，LongBenchPro/LongBench与Zipf workload20%sharedprefix；crossdocument reasoning明显局限，cachehit/TTFT为servingloop。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13060v1#S2", "https://arxiv.org/html/2610.13060v1#S4", "https://arxiv.org/html/2610.13060v1#Sx1" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13060v1", "sha256": "b314b98efc2856dcd457c925492d574ccc6b86634649118f6422555ab3c7abdf" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13058", "title": "Travel Back To Move Forward: Retrospective Exploration and Evolving for Long Horizon Agents", "published": "2026-10-09T06:11:39Z", "reviews": [ { "id": "2610.13058", "title": "Travel Back To Move Forward: Retrospective Exploration and Evolving for Long Horizon Agents", "published": "2026-10-09T06:11:39Z", "decision": "reviewed_deferred", "reason": "ReTRAIL真实恢复环境状态并保留abandoned exploration经验，以State-Anchored GRPO与branch advancement reward共同训练；MineSweeper/Sokoban/ALFWorld/ScienceWorld评测，具Facebook署名线索优先审查，需真实rewind接口，冻结16外下一批。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Retrospective Exploration", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Learning from Retrospective Exploration", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13058v1", "sha256": "dac2da4a6c200b7b10b09a96b67249a22ccc39e56dfe33189fffc5e86dea9c57", "local_html_sha256": "dac2da4a6c200b7b10b09a96b67249a22ccc39e56dfe33189fffc5e86dea9c57", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13058v1", "result": "retrieved", "sha256": "dac2da4a6c200b7b10b09a96b67249a22ccc39e56dfe33189fffc5e86dea9c57" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.13057", "title": "CoBe: A Benchmark For Conversational Counterfactual Text Editing", "published": "2026-10-09T06:11:12Z", "reviews": [ { "id": "2610.13057", "title": "CoBe: A Benchmark For Conversational Counterfactual Text Editing", "published": "2026-10-09T06:11:12Z", "decision": "reviewed_deferred", "reason": "Columbia CoBe 2000题counterfactual text editing与保持因果无关事实的scorer；§3构造与§4 nine-model评测可作LLM benchmark候选，须核公开样本与评分器可获取性后接入。没有新工业搜广推算法，留下一批基础模型评测。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "bounded_methods_and_evaluation", "review_artifact": "research-audits/oct12-review-closure/recommendation-reviews.json", "evidence_locations": [ "§3 CoBe Dataset", "§4 Results", "§5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13057v1", "sha256": "b1d6cf3e0ecc0d2733bd27ed9504c7fe7283f00e71ffd445dc395c2f2c36d7bd" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "recommendation" ], "status": "reviewed_deferred" }, { "id": "2610.13056", "title": "A4-MR: Abnormality-aware Anatomical Alignment for MR", "published": "2026-10-09T06:10:10Z", "reviews": [ { "id": "2610.13056", "title": "A4-MR: Abnormality-aware Anatomical Alignment for MR", "published": "2026-10-09T06:10:10Z", "decision": "reviewed_deferred", "reason": "A4MR 3Dvisionlanguagepretrain anatomicalnoisylocalization/image-textregion对齐，healthy同organ可share positives，abnormal同organ不同finding区分；不是普通wholevolumeCLIP。; 约50kMRIstudies/491kseries，下游frozenencoder localization/segmentation/pathologytasks；MedSAM2带gtboxes比较属于额外oraclecontext。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13056v1#S3", "https://arxiv.org/html/2610.13056v1#S4", "https://arxiv.org/html/2610.13056v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13056v1", "sha256": "54f90c196665df175135b06eefb00a4d14a3e07c321446cd75902939e7d476bb" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13045", "title": "Redirecting Teacher Supervision: Temporal-Response-Guided Latent Distillation for Video-Language Models", "published": "2026-10-09T06:01:48Z", "reviews": [ { "id": "2610.13045", "title": "Redirecting Teacher Supervision: Temporal-Response-Guided Latent Distillation for Video-Language Models", "published": "2026-10-09T06:01:48Z", "decision": "reviewed_deferred", "reason": "TRG在acceptedteacherresponse固定context下用student局部temporalperturb/JVP response构orthoprojectorP；teacher correction δ→δ+Pδ再restore原norm，保留span外分量，detachedtarget只redirectdirection。; VideoR1SFT+OPD、4B/9Bstudent27Bteacher四bench；与PlainAS/randomorientation对比，VideoMMMU多数来自plainalignment。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13045v1#S3", "https://arxiv.org/html/2610.13045v1#S4", "https://arxiv.org/html/2610.13045v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13045v1", "sha256": "ad3f12364320a3e8d42d807abd6c73c427767a5a79477f49f9b1db5aee810697" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13041", "title": "RAMP: Recurrent Adaptive Mixed Precision for Efficient Prefill in Recurrent-Depth Language Models", "published": "2026-10-09T05:57:33Z", "reviews": [ { "id": "2610.13041", "title": "RAMP: Recurrent Adaptive Mixed Precision for Efficient Prefill in Recurrent-Depth Language Models", "published": "2026-10-09T05:57:33Z", "decision": "reviewed_deferred", "reason": "同一shared recurrent block前K步W+A低precision、terminal BF16 cleanup；fixedT/matched precisionbudget逆序和随机schedule对照，理论依赖local contraction。; §6 Huginn0125 T32/K24，frozen nested-context WT2/PG19 identicaltail64 targets；H100 nativeFP8 first-prefill同optimizedemptycache路径。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract/author information/main defining methods/main experiments and tables; production-evidence paragraph spot check; appendices not exhaustively read", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13041v1#S2", "https://arxiv.org/html/2610.13041v1#S3", "https://arxiv.org/html/2610.13041v1#S4", "https://arxiv.org/html/2610.13041v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13041v1", "sha256": "1eba41d730c3d49c74e8ee6eecde3c93cbb663950961faa547faafa9222d97cf" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.13040", "title": "Lightweight Time-Frequency Vibration Sensing for Consumer Device Condition Monitoring", "published": "2026-10-09T05:57:15Z", "reviews": [ { "id": "2610.13040", "title": "Lightweight Time-Frequency Vibration Sensing for Consumer Device Condition Monitoring", "published": "2026-10-09T05:57:15Z", "decision": "rejected_or_out_of_scope", "reason": "CWT/HHT将振动变timefrequency图，再MobileNet/ConvNext/YOLO11 classifier；区别randomwindowsplit与bearing/timeblockstructuredtrain/val/test，记录preprocess+CPU网络成本。; CWRU/PU/IMS bearingbench和i7-13700K batch1，非真实consumerappliance/embeddedMCU部署。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13040v1#S3", "https://arxiv.org/html/2610.13040v1#S4", "https://arxiv.org/html/2610.13040v1#S5", "https://arxiv.org/html/2610.13040v1#S7" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13040v1", "sha256": "10e04f2196460eee849f13963691ceffaa67e9283c8dd5d0f5771e3e31e9547b" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.13035", "title": "GeoGR: Spatial-Semantic Token Compression and Group Routing for Efficient 3D VLMs", "published": "2026-10-09T05:46:54Z", "reviews": [ { "id": "2610.13035", "title": "GeoGR: Spatial-Semantic Token Compression and Group Routing for Efficient 3D VLMs", "published": "2026-10-09T05:46:54Z", "decision": "reviewed_deferred", "reason": "GeoGR voxel GeoSemZip与LLM recoverable GroupRoute结合compression-aware 67k posttraining，ScanNet五benchmark/真实prefill评测；foundation3D VLM下一批需RGB/depth/pose和可恢复token状态，不用文本mask proxy。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "4 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13035v1", "sha256": "8c0d8440bbe5b6cc81360366e17d139bd3312b70d0680b777a432e06a369fcb6", "local_html_sha256": "8c0d8440bbe5b6cc81360366e17d139bd3312b70d0680b777a432e06a369fcb6", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13035v1", "result": "retrieved", "sha256": "8c0d8440bbe5b6cc81360366e17d139bd3312b70d0680b777a432e06a369fcb6" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.13025", "title": "Evo$^2$: Evolving the Evolver through Weight-Harness Co-Evolution", "published": "2026-10-09T05:28:02Z", "reviews": [ { "id": "2610.13025", "title": "Evo$^2$: Evolving the Evolver through Weight-Harness Co-Evolution", "published": "2026-10-09T05:28:02Z", "decision": "accepted", "reason": "Google DeepMind/UIUC冻结16范围已有Evo2实现：shared Proposer/Reflector真实SFT+interleaved role GRPO、fresh counterpart、executor固定、实执行树编辑/交叉与恢复。原文已核；本地prompt-only harness是concept_demo，GPU诊断与OOM由root记录，非能力复现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 14068 ] }, { "heading": "4 Experiments", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 8725 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13025v1", "sha256": "0bdabf2558023e54f2a2da398c1d385b48834096451d775312b2f0432cb20d2c", "local_html_sha256": "0bdabf2558023e54f2a2da398c1d385b48834096451d775312b2f0432cb20d2c", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13025v1", "result": "retrieved", "sha256": "0bdabf2558023e54f2a2da398c1d385b48834096451d775312b2f0432cb20d2c" } ] }, "new_implementation_inferred_from_review": false }, { "id": "2610.13025", "title": "Evo$^2$: Evolving the Evolver through Weight-Harness Co-Evolution", "published": "2026-10-09T05:28:02Z", "decision": "accepted", "reason": "用户点名冻结16范围；selected defining-method/experiment原文证据核验，具体实现/运行等级独立记录。; 共享Proposer/Reflector通过真实源编辑、交叉与轨迹前缀恢复进化harness；warmup SFT后交替角色GRPO，每轮fresh冻结counterpart，executor固定；候选由实际evaluator反馈树选择。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "实施期间读取摘要、中心方法/训练/评测及部分相关附录；本次复用这些已读来源并重新校验官方HTML字节SHA和v1日期，不声明全文每段已读。", "review_artifact": "research-audits/oct12-review-closure/frozen-agent-reviews.json", "evidence_locations": [ "3.1 Overview", "3.2 Harness evolution through dual-role self-modification", "3.3.1 Data collection", "3.3.2 Reward design", "3.3.3 Training process", "4.1 Experimental Settings", "4.2 Self-Evolution on Coding Benchmarks", "4.3 Harness Transfer across Benchmarks", "4.4 Self-Evolution with Proprietary Models", "4.5 Self-Evolution on Memory Tasks", "4.6 In-Depth Analysis" ], "dateproof": { "url": "https://arxiv.org/abs/2610.13025", "sha256": "7a1d0eb105e3fcd0c4f9030aaf1af50e91a4388b2ea7cda66ec70b0bc59c4fe8", "location": "Submission history / [v1]", "first_publication_utc": "2026-10-09T05:28:02Z" }, "source_proof": { "url": "https://arxiv.org/html/2610.13025v1", "sha256": "0bdabf2558023e54f2a2da398c1d385b48834096451d775312b2f0432cb20d2c", "byte_hash_verified_on": "2026-10-12" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "accepted", "implementation_key": "evo2", "accepted_scope": "bounded faithful mechanism/runtime or evaluation infrastructure; not original capability replication" }, { "id": "2610.13022", "title": "Don't Guess, Just Ask: Evaluating Interactive Requirement Clarification for Repository-Level Software Development", "published": "2026-10-09T05:23:01Z", "reviews": [ { "id": "2610.13022", "title": "Don't Guess, Just Ask: Evaluating Interactive Requirement Clarification for Repository-Level Software Development", "published": "2026-10-09T05:23:01Z", "decision": "reviewed_deferred", "reason": "已读ReqClarifyBench交互设计与评测：341 repository任务，仅vague intent进入被评模型，persona user按真实需求回应；clarification效果与最终需求文档质量不同。下一批需隔离user私有spec与外部code evaluation。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4. Experiment Design", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5. Evaluation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13022v1", "sha256": "d97ad7e4ad7cb7f8f2702a7bb960755f6b45f0dd4ebeeaccf96090e989ae012f", "local_html_sha256": "d97ad7e4ad7cb7f8f2702a7bb960755f6b45f0dd4ebeeaccf96090e989ae012f", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.13022v1", "result": "retrieved", "sha256": "d97ad7e4ad7cb7f8f2702a7bb960755f6b45f0dd4ebeeaccf96090e989ae012f" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.13021", "title": "EmbodiedCLBench: Evaluating Continual Learning for Self-Evolving Embodied Agents", "published": "2026-10-09T05:23:00Z", "reviews": [ { "id": "2610.13021", "title": "EmbodiedCLBench: Evaluating Continual Learning for Self-Evolving Embodied Agents", "published": "2026-10-09T05:23:00Z", "decision": "reviewed_deferred", "reason": "TongSim构object-action任务/stategraph，basicexperience→advancedcompositionalunseen任务；offline交互选择含optimal/suboptimal/invalid可执行action候选的确定性benchmark。; 168basic+100advanced，ICL/skilllearning/agentharness对照；更多experience不持续提升，宽stepbudget也不替代经验。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.13021v1#S3", "https://arxiv.org/html/2610.13021v1#S4", "https://arxiv.org/html/2610.13021v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.13021v1", "sha256": "144ba4b449d1ff386c07485bd0b5349298baf100174b7481b80495cf50f0b760" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12998", "title": "Evaluating LLM Creativity through Long-Tail Performance", "published": "2026-10-09T05:03:18Z", "reviews": [ { "id": "2610.12998", "title": "Evaluating LLM Creativity through Long-Tail Performance", "published": "2026-10-09T05:03:18Z", "decision": "reviewed_deferred", "reason": "固定reference输入输出分布，以Pile ngramabsence novelty选existingbenchmarktailfraction，q@Tail(alpha)为tail任务quality均值；对LM-Arena humancreativewritingrank做Kendall对照。; 15bench/26openmodels，tailness与rank相关；多数creativitypromoting方法提升peaknovelty却降低tailquality。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12998v1#S3", "https://arxiv.org/html/2610.12998v1#S4", "https://arxiv.org/html/2610.12998v1#S6", "https://arxiv.org/html/2610.12998v1#Sx1" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12998v1", "sha256": "53815f132bc3b1eb5354a2d6a043431af78938b84de3329b9d2d7a43d5dc34e7" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12995", "title": "What Should SFT Learn? Conditional Information Gain for Generalizable Post-Training", "published": "2026-10-09T04:57:55Z", "reviews": [ { "id": "2610.12995", "title": "What Should SFT Learn? Conditional Information Gain for Generalizable Post-Training", "published": "2026-10-09T04:57:55Z", "decision": "reviewed_deferred", "reason": "CIG-SFT frozen reference有/无task条件likelihood在gold训练prefix上估计conditional information gain，用KL projection等价weighted SFT；math/code公开任务与baseline，离线train supervision合法但gold不得进入被评估rollout，下一批。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12995v1", "sha256": "2a7ba2a327ec42e00bb6dc364e69b71fad8d09a7b773b921b72d262c21b0d374", "local_html_sha256": "2a7ba2a327ec42e00bb6dc364e69b71fad8d09a7b773b921b72d262c21b0d374", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12995v1", "result": "retrieved", "sha256": "2a7ba2a327ec42e00bb6dc364e69b71fad8d09a7b773b921b72d262c21b0d374" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.12981", "title": "The Evidence Anchoring Boundary: Where Visual Interventions Matter and When They Help in Multimodal LLMs", "published": "2026-10-09T04:26:50Z", "reviews": [ { "id": "2610.12981", "title": "The Evidence Anchoring Boundary: Where Visual Interventions Matter and When They Help in Multimodal LLMs", "published": "2026-10-09T04:26:50Z", "decision": "reviewed_deferred", "reason": "depthmatchedRMS visual-residual edit定位operationalEAB；counterfactualsubtraction按presence/attribute/relation calibrationselective routing，保留原预测除非calibration支持该类型干预。; 六checkpoint三family，EAB frozen oncalibration+disjointeval；attribute五折image-disjoint routing，小提升而universal/跨calibratedscope可能harm。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12981v1#S3", "https://arxiv.org/html/2610.12981v1#S4", "https://arxiv.org/html/2610.12981v1#S5", "https://arxiv.org/html/2610.12981v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12981v1", "sha256": "4e7681c86215ea3e9eb89efbc3269de67bc142256618a1d6aa994ad3fbdb0fec" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12975", "title": "SmartSalesBench: A Resistance-Aware Multi-Turn Benchmark for Evaluating LLM Sales Agents", "published": "2026-10-09T04:21:12Z", "reviews": [ { "id": "2610.12975", "title": "SmartSalesBench: A Resistance-Aware Multi-Turn Benchmark for Evaluating LLM Sales Agents", "published": "2026-10-09T04:21:12Z", "decision": "reviewed_deferred", "reason": "已读SmartSalesBench设计/指标：抵抗buyer personas的模拟20轮会话，RAG产品、13680session、50人工judge校准；探索指标尚待报告。下一批审查保留，但小人工校准与模拟转化率不能声称真实sales A/B，不优先实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "III Benchmark Design", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "IV Evaluation Framework", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12975v1", "sha256": "78ec2fc4a5e8f26b5c3c62302df7f2a8d8d15c8a79dfdddc4194883a4af1bf40", "local_html_sha256": "78ec2fc4a5e8f26b5c3c62302df7f2a8d8d15c8a79dfdddc4194883a4af1bf40", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12975v1", "result": "retrieved", "sha256": "78ec2fc4a5e8f26b5c3c62302df7f2a8d8d15c8a79dfdddc4194883a4af1bf40" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12961", "title": "MVGraph: Benchmarking LVLMs on Evidence Composition across Multi-View Visual Graph", "published": "2026-10-09T03:59:34Z", "reviews": [ { "id": "2610.12961", "title": "MVGraph: Benchmarking LVLMs on Evidence Composition across Multi-View Visual Graph", "published": "2026-10-09T03:59:34Z", "decision": "reviewed_deferred", "reason": "MVGraph程序生成/验证partial/temporal多graphview确需多视图，真实渲染输入；8maingraphtasks+29stage perception/alignment/composition/reasoning诊断。; 8kmain+64kdiagnosticrecords，15LVLM；mainendtoend和progressivegoldintermediateoracle定位瓶颈，dropviews需保序temporal约束。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12961v1#S3", "https://arxiv.org/html/2610.12961v1#S4", "https://arxiv.org/html/2610.12961v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12961v1", "sha256": "2a42bfe82aa3af3631073b9a0f8f739781f89f4f5c0d7781be86bcc3fb07d793" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12959", "title": "Video Codec-Guided Training-Free Token Compression for Video Large Language Models", "published": "2026-10-09T03:56:58Z", "reviews": [ { "id": "2610.12959", "title": "Video Codec-Guided Training-Free Token Compression for Video Large Language Models", "published": "2026-10-09T03:56:58Z", "decision": "reviewed_deferred", "reason": "真实codec referenceframe/motionvector/residual对齐token grid；residual只生成protection候选池，referencechainmotion只限定跨sampledframe merge区域，最终仍localvisualfeaturesimilarity决定。; 三VLLM四videobench对FlashVID，主要减少feature comparisons；residual相邻变化加权与阈值有消融。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12959v1#S3", "https://arxiv.org/html/2610.12959v1#S4", "https://arxiv.org/html/2610.12959v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12959v1", "sha256": "60efeb55686c481409088738317fb82cef5a6fbf7e593a4bd29a16e4e0bb4ee3" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12957", "title": "JoyInside-VLM-Agent: Product-Conditioned Multimodal Policy Learning for Embodied Agents", "published": "2026-10-09T03:54:58Z", "reviews": [ { "id": "2610.12957", "title": "JoyInside-VLM-Agent: Product-Conditioned Multimodal Policy Learning for Embodied Agents", "published": "2026-10-09T03:54:58Z", "decision": "reviewed_deferred", "reason": "已读JoyInside方法/实验：产品条件multimodal trajectory、SFT、多teacher OPD及DPER token-level privileged evidence路由；部署单student不读teacher reference。私有120k混合数据、领域skill及两机器人case限制公开复现，下一批须保留真实多阶段训练。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12957v1", "sha256": "47cbf6327b542d9dba8e3dcc4425451ca35d20b13593505db8603e4d3d12ea89", "local_html_sha256": "47cbf6327b542d9dba8e3dcc4425451ca35d20b13593505db8603e4d3d12ea89", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12957v1", "result": "retrieved", "sha256": "47cbf6327b542d9dba8e3dcc4425451ca35d20b13593505db8603e4d3d12ea89" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12942", "title": "In-Context Learning Can Help Vision-Language Models Overcome Training Priors", "published": "2026-10-09T03:40:07Z", "reviews": [ { "id": "2610.12942", "title": "In-Context Learning Can Help Vision-Language Models Overcome Training Priors", "published": "2026-10-09T03:40:07Z", "decision": "reviewed_deferred", "reason": "visualICL controlledprior-aligned/conflicting counterfactual任务，complex/simple/no-prior settings区分formatlearning、prioroverride与底层感知；同demonstration/promptnormalize跨模型。; 五visiontasks五VLMfamilies，机制attention/masking分析；highcardinalitycounting不稳定、部分10/15count低于zeroshot。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12942v1#S3", "https://arxiv.org/html/2610.12942v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12942v1", "sha256": "839729f5e226c755936005ff396f178f492a054fe7b3897114f08584b823494b" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12932", "title": "SEEDS: Preview-Guided Evidence Admission for Search Agents", "published": "2026-10-09T03:25:39Z", "reviews": [ { "id": "2610.12932", "title": "SEEDS: Preview-Guided Evidence Admission for Search Agents", "published": "2026-10-09T03:25:39Z", "decision": "reviewed_deferred", "reason": "已读SEEDS方法/实验：retrieved preview后由当前reasoning状态选择任意subset/cardinality才admit full evidence，local corpus/live web九QAbench。下一批必须真实检索和模型selection，不能固定rankprefix或gold evidence替代；live search本身不等于线上A/B。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12932v1", "sha256": "829d1c6dc5b2c2356ee5d637a5ade91508ed2b4575811a6bf665a0daea7ffaaf", "local_html_sha256": "829d1c6dc5b2c2356ee5d637a5ade91508ed2b4575811a6bf665a0daea7ffaaf", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12932v1", "result": "retrieved", "sha256": "829d1c6dc5b2c2356ee5d637a5ade91508ed2b4575811a6bf665a0daea7ffaaf" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12927", "title": "Collaborative Learning and Reasoning over Model Slices", "published": "2026-10-09T03:17:29Z", "reviews": [ { "id": "2610.12927", "title": "Collaborative Learning and Reasoning over Model Slices", "published": "2026-10-09T03:17:29Z", "decision": "reviewed_deferred", "reason": "已读model slicing/pyramid supervision与实验：可比总parameter budget的异构小模型，protocol distillation、逐步score与final reward共同训练collaboration，四任务五bench。下一批需真正构建/训练多model，非prompt角色切换。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12927v1", "sha256": "b65ffa2825129a5464c5e61092be38e96cb5532c7adacd8cc24671832d05f3a3", "local_html_sha256": "b65ffa2825129a5464c5e61092be38e96cb5532c7adacd8cc24671832d05f3a3", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12927v1", "result": "retrieved", "sha256": "b65ffa2825129a5464c5e61092be38e96cb5532c7adacd8cc24671832d05f3a3" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12907", "title": "BLEND: Balancing Personalization vs. Generalization in Federated Vision-Language Models", "published": "2026-10-09T02:57:18Z", "reviews": [ { "id": "2610.12907", "title": "BLEND: Balancing Personalization vs. Generalization in Federated Vision-Language Models", "published": "2026-10-09T02:57:18Z", "decision": "reviewed_deferred", "reason": "BLEND CLIP每layerglobalvision/textadapters跨clientaggregate，personalvisionprojection留local；blendglobal/personaloutputs并用frozenzeroshotprojectionheads anchor促进unseen泛化。; 10client simulatedFL、多rank/layer/γλsweep、3L40S；personalization/generalizationHmean tradeoff。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12907v1#S3", "https://arxiv.org/html/2610.12907v1#S4", "https://arxiv.org/html/2610.12907v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12907v1", "sha256": "33f117ec6aaaad69392d054b843a165984badf4f1e3c56098d96ce15fa52f11a" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12885", "title": "Personalized Federated Direct Preference Optimization via LoRA Adapters", "published": "2026-10-09T02:20:29Z", "reviews": [ { "id": "2610.12885", "title": "Personalized Federated Direct Preference Optimization via LoRA Adapters", "published": "2026-10-09T02:20:29Z", "decision": "reviewed_deferred", "reason": "pFedDPO同步global base、private LoRA保留client preference，SHP/hh5 clients非IID实验、主要gradient convergence；下一批需真实FedAvg与local adapters，private数据不能centralize，训练loss不等于user效用。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "Personalized FedDPO (pFedDPO)", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12885v1", "sha256": "b6753f95bca1a75347d72e607738bd7b4248bdaaf6fa7123fc7a67b52a2dae62", "local_html_sha256": "b6753f95bca1a75347d72e607738bd7b4248bdaaf6fa7123fc7a67b52a2dae62", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12885v1", "result": "retrieved", "sha256": "b6753f95bca1a75347d72e607738bd7b4248bdaaf6fa7123fc7a67b52a2dae62" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.12874", "title": "Temporal Evidence Forests for Long-Horizon Multi-Party Conversational Memory", "published": "2026-10-09T02:05:38Z", "reviews": [ { "id": "2610.12874", "title": "Temporal Evidence Forests for Long-Horizon Multi-Party Conversational Memory", "published": "2026-10-09T02:05:38Z", "decision": "reviewed_deferred", "reason": "原聊天存储不改写；metadata-expanded BM25种子、DBSCAN时间局部证据发现、Update/Expand证据forest和保序linearization组合，回答模型不见reference。; GroupMemBench/EverMemBench官方长程多人对话；topic/phase移除消融仍保留部分优势。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12874v1#S4", "https://arxiv.org/html/2610.12874v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12874v1", "sha256": "cc0ee64ed538c269d993d0b91cf56e97bbbb657ef48541df13375a87a7d5ab52" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12866", "title": "CaptchaAgent: Verifiable Visual Reasoning for CAPTCHA Solving", "published": "2026-10-09T01:50:26Z", "reviews": [ { "id": "2610.12866", "title": "CaptchaAgent: Verifiable Visual Reasoning for CAPTCHA Solving", "published": "2026-10-09T01:50:26Z", "decision": "rejected_or_out_of_scope", "reason": "研究structuredvisualstate与一致性诊断对有限视觉puzzle的作用；本审查只保留高层视觉推理/防御研究信息，不形成CAPTCHA自动破解实施方案。; OpenCaptchaWorld20固定family browserexecutedresearchbenchmark与Oedipus小offlinejudge任务；后者非同一Pass1定义，支出只模型调用。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12866v1#S3", "https://arxiv.org/html/2610.12866v1#S4", "https://arxiv.org/html/2610.12866v1#S5", "https://arxiv.org/html/2610.12866v1#S6", "https://arxiv.org/html/2610.12866v1#S8" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12866v1", "sha256": "b12f862d37801643ffe11359c2d57c4db4a87ccd2f05b17623fde3b2c5bc1b93" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.12864", "title": "Time Delay System for Sequence Modeling", "published": "2026-10-09T01:43:37Z", "reviews": [ { "id": "2610.12864", "title": "Time Delay System for Sequence Modeling", "published": "2026-10-09T01:43:37Z", "decision": "reviewed_deferred", "reason": "TDS semanticlagsoftattention取历史context/similarity，diagonalstate用closedformcumsum，TSKfuzzyprototype rules调instant与delayupdate；不是普通LSTMhiddenstate。; IMDB分类、CNNDailyMail80ktrain+2kval/test文本生成，H100；仅两任务及lag数敏感性。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12864v1#S3", "https://arxiv.org/html/2610.12864v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12864v1", "sha256": "3a30d2163a993f87a82fb9fd9d58329c172781fe698bd7aaa87d430045488caa" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12849", "title": "Co-Modifying Application Code and Deployment Infrastructure: A Causal Study of AI Agent Pull Requests", "published": "2026-10-09T01:16:04Z", "reviews": [ { "id": "2610.12849", "title": "Co-Modifying Application Code and Deployment Infrastructure: A Causal Study of AI Agent Pull Requests", "published": "2026-10-09T01:16:04Z", "decision": "reviewed_deferred", "reason": "已读PR观察研究设计/结果：26999 agent/human PR、四档path taxonomy与4D exact strata matching，在conditional exchangeability下估计双scope集成overhead；真实repo数据是工业实证但不是随机A/B，未观测混杂仍待检验。下一批评测infra，不自动扩大部署权限。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2. Empirical Study Design", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3. Empirical Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12849v1", "sha256": "ecdaabac47c3b913f2826d9a975bfc20ffb63762eb318135b41e502c7dae8a8b", "local_html_sha256": "ecdaabac47c3b913f2826d9a975bfc20ffb63762eb318135b41e502c7dae8a8b", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12849v1", "result": "retrieved", "sha256": "ecdaabac47c3b913f2826d9a975bfc20ffb63762eb318135b41e502c7dae8a8b" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12822", "title": "Reward Hacking as a Transmissible Strategy in Multi-Agent Coding Systems", "published": "2026-10-09T00:02:59Z", "reviews": [ { "id": "2610.12822", "title": "Reward Hacking as a Transmissible Strategy in Multi-Agent Coding Systems", "published": "2026-10-09T00:02:59Z", "decision": "reviewed_deferred", "reason": "Meta优先。已读问题定义、单agent源、角色传播与缓解段：用leaky测试训练出的reward hacks跨代码/共享记忆/规划传播，独立strict executor和静态核验区分真实hack。是安全评测研究，不是可宣称提升的agent算法；生成恶意测试不得无权运行在宿主。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Problem Formulation", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 4428 ] }, { "heading": "4 Single-Agent Reward Hacking and Source Agents", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 4584 ] }, { "heading": "6 Role Transfer: Planner-to-Coder Propagation", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 6633 ] }, { "heading": "8 Mitigation", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 2079 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12822v1", "sha256": "598c47ec103ac31696840c61c0ba49bd24b6da80ea03d8c29e2bd50804a295fb", "local_html_sha256": "598c47ec103ac31696840c61c0ba49bd24b6da80ea03d8c29e2bd50804a295fb", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12822v1", "result": "retrieved", "sha256": "598c47ec103ac31696840c61c0ba49bd24b6da80ea03d8c29e2bd50804a295fb" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12813", "title": "VEPA: Verified Evolving Formal Policy for Lifelong LLM Agent Guardrails", "published": "2026-10-08T23:49:50Z", "reviews": [ { "id": "2610.12813", "title": "VEPA: Verified Evolving Formal Policy for Lifelong LLM Agent Guardrails", "published": "2026-10-08T23:49:50Z", "decision": "reviewed_deferred", "reason": "Google优先。已读VEPA方法缓存及官方§3.2.2权限段、实验：typed DSL与确定性wellformed/conflict校验、Pareto修复/回归archive、逆行为密度采样，Beta后验误差门槛决定override authority。五seed prequential benchmark流不是生产A/B；auto parsing语义之外不受形式保证。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 16000 ] }, { "heading": "4 Experiments", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 9263 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12813v1", "sha256": "a69f3cc9c9d8d542cd5c20ec6a877f0e12f0cd8d606979e6028f43bdfb168748", "local_html_sha256": "a69f3cc9c9d8d542cd5c20ec6a877f0e12f0cd8d606979e6028f43bdfb168748", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12813v1", "result": "retrieved", "sha256": "a69f3cc9c9d8d542cd5c20ec6a877f0e12f0cd8d606979e6028f43bdfb168748" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12794", "title": "Meta-Reflective Evolution Effectively Improves Production AI Systems", "published": "2026-10-08T23:00:44Z", "reviews": [ { "id": "2610.12794", "title": "Meta-Reflective Evolution Effectively Improves Production AI Systems", "published": "2026-10-08T23:00:44Z", "decision": "reviewed_deferred", "reason": "已读CoCoEvolve方法与实验协议：archive parent、history检索、live target probing、自测mutation再external评分；train实例可见，validation仅score、test隔离。production AI目标及公开heldout基准不自动证明生产流量A/B，下一批须进一步核验live authority和真实evaluator，非registry标签。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 CoCoEvolve", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12794v1", "sha256": "21dfb86db8f99a2613a96dba44dd9470b221a554775213c0ddd31c25f768237f", "local_html_sha256": "21dfb86db8f99a2613a96dba44dd9470b221a554775213c0ddd31c25f768237f", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12794v1", "result": "retrieved", "sha256": "21dfb86db8f99a2613a96dba44dd9470b221a554775213c0ddd31c25f768237f" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12792", "title": "Large Language Models, Machine-Assisted Cultural Analytics, and the Agentic Turn", "published": "2026-10-08T22:54:57Z", "reviews": [ { "id": "2610.12792", "title": "Large Language Models, Machine-Assisted Cultural Analytics, and the Agentic Turn", "published": "2026-10-08T22:54:57Z", "decision": "rejected_or_out_of_scope", "reason": "已读摘要及annotation reliability/quantitization段：文化analytics的论证性章节，讨论测量与解释不确定性，不提供独立agent算法或可执行benchmark，范围外背景方法参考。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 Annotation Reliability and Modeling Uncertainty", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "1 LLM-Assisted Quantitization, Annotation and Classification", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12792v1", "sha256": "848337300a788ca337913f4a1bea15e47cde0a0aee932be9223a2bcde9666de4", "local_html_sha256": "848337300a788ca337913f4a1bea15e47cde0a0aee932be9223a2bcde9666de4", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12792v1", "result": "retrieved", "sha256": "848337300a788ca337913f4a1bea15e47cde0a0aee932be9223a2bcde9666de4" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.12789", "title": "EvoBelief: Falsifiable World Models for Self-Improving LLM Agents", "published": "2026-10-08T22:46:14Z", "reviews": [ { "id": "2610.12789", "title": "EvoBelief: Falsifiable World Models for Self-Improving LLM Agents", "published": "2026-10-08T22:46:14Z", "decision": "reviewed_deferred", "reason": "已读EvoBelief方法/实验：falsifiable hypothesis按异常/未知形成，经独立observation/controlled trials验证后限定scope，counterexample缩限或退休，LM frozen。ALFWorld/ScienceWorld/MiniGrid/DiscoveryWorld公开环境；下一批不能把真dynamics当agent输入或省掉验证闭环。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12789v1", "sha256": "34286d36ae54f1878e5b29974df5c4cfd268e8eac021a00c3031f17babb21f62", "local_html_sha256": "34286d36ae54f1878e5b29974df5c4cfd268e8eac021a00c3031f17babb21f62", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12789v1", "result": "retrieved", "sha256": "34286d36ae54f1878e5b29974df5c4cfd268e8eac021a00c3031f17babb21f62" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12783", "title": "Shift-Aware Identity-Guided Latent Refinement for Referring Audio--Visual Segmentation", "published": "2026-10-08T22:39:41Z", "reviews": [ { "id": "2610.12783", "title": "Shift-Aware Identity-Guided Latent Refinement for Referring Audio--Visual Segmentation", "published": "2026-10-08T22:39:41Z", "decision": "reviewed_deferred", "reason": "SAIL tri-modalRefAVS的frame-dynamictoken与globalsemanticanchor分离，shiftawarelatentrefinement按domain证据纠漂移，SAM2promptdecoder+连续memory保持referentidentity。; RefAVSBench officialvideo split含seen/unseen/null与R2AVSBench；目标变化/遮挡/声音衰弱是主要诊断情形。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12783v1#S3", "https://arxiv.org/html/2610.12783v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12783v1", "sha256": "cd24a584bd27fdb09e655b05c9c5f50a4888f46f4c2821d28d623d17db7e2fa1" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12757", "title": "MADS: Multi-Agent Dialog Simulation for Robust Conversational Agents", "published": "2026-10-08T22:08:10Z", "reviews": [ { "id": "2610.12757", "title": "MADS: Multi-Agent Dialog Simulation for Robust Conversational Agents", "published": "2026-10-08T22:08:10Z", "decision": "reviewed_deferred", "reason": "MADS三agent模拟user/assistant/tool含adversarial personas并验证dialogue，SFT+RLVR训练robust tool dialogs；下一批需检查106 training trajectories与1095总conversations split isolation，真实tool验证避免泄漏。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 MADS-Bench", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12757v1", "sha256": "053fe9f848f542a6258a943471764084edddcd5ba9f07284da03682037c146f0", "local_html_sha256": "053fe9f848f542a6258a943471764084edddcd5ba9f07284da03682037c146f0", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12757v1", "result": "retrieved", "sha256": "053fe9f848f542a6258a943471764084edddcd5ba9f07284da03682037c146f0" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.12734", "title": "StereoEngine: Scaling Stereo Bimanual Robot Data without Stereo Cameras", "published": "2026-10-08T21:28:58Z", "reviews": [ { "id": "2610.12734", "title": "StereoEngine: Scaling Stereo Bimanual Robot Data without Stereo Cameras", "published": "2026-10-08T21:28:58Z", "decision": "reviewed_deferred", "reason": "StereoEngineL monocularvideo depthwarp生成rightview后videodiffusion inpaintdisocclusion；S从simstereo联合生成photo-real左右，无realvideo输入；actualstereodata用于policy midtrain/finetune。; 两PiPERarms/ZEDMini2 realrollouts，3469公共monoepisodes各baseline同源转换；下游context实验仅proprio无languageinstruction。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12734v1#S3", "https://arxiv.org/html/2610.12734v1#S4", "https://arxiv.org/html/2610.12734v1#S5", "https://arxiv.org/html/2610.12734v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12734v1", "sha256": "0698a5d5dc301554f655f3660efe5b456d139762d9188884f9af4c385d6935fe" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12711", "title": "Byzantine-Resilient Federated Large Language Model via Hypergraph Message Passing", "published": "2026-10-08T20:47:08Z", "reviews": [ { "id": "2610.12711", "title": "Byzantine-Resilient Federated Large Language Model via Hypergraph Message Passing", "published": "2026-10-08T20:47:08Z", "decision": "reviewed_deferred", "reason": "FedLLM frozenbase+LoRA clientupdates作为hypergraphnodes，HMP+GAE结合历史checkpoint关系/reconstruction检测和降权不一致更新；server无clientrawdata、benign身份未知。; Qwen2.5/Llama3.2 AGNews/YahooAnswers分类，RTXPRO6000Blackwell；CSE/PPL与accuracy比较GAE及多defenses。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12711v1#S3", "https://arxiv.org/html/2610.12711v1#S4", "https://arxiv.org/html/2610.12711v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12711v1", "sha256": "1fe8a1589810401a3c795fa6aa1718324c28136d87fbe9f5926f37e538737498" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12707", "title": "Unsupervised Multimodal Reward Modeling", "published": "2026-10-08T20:38:25Z", "reviews": [ { "id": "2610.12707", "title": "Unsupervised Multimodal Reward Modeling", "published": "2026-10-08T20:38:25Z", "decision": "reviewed_deferred", "reason": "冻结Qwen3-VL intermediate-layer linear head用same-image caption perturbations自监督，标签只用于layer选择；reward benchmark与verifier-driven bootstrap downstream不等价，下一批需train/val选择层隔离和真实multimodal heads。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "2 Methodology", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12707v1", "sha256": "bf8fe4d9ecf041ecb1a8efb2caee64631527f1aa5627e740cfaeca358e8e6eb8", "local_html_sha256": "bf8fe4d9ecf041ecb1a8efb2caee64631527f1aa5627e740cfaeca358e8e6eb8", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12707v1", "result": "retrieved", "sha256": "bf8fe4d9ecf041ecb1a8efb2caee64631527f1aa5627e740cfaeca358e8e6eb8" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false }, { "id": "2610.12707", "title": "Unsupervised Multimodal Reward Modeling", "published": "2026-10-08T20:38:25Z", "decision": "accepted", "reason": "frozenQwen3VL中间language-layer responsefeatures上的biasfreelinearrewardhead，captionnear/span/splice/shuffle等perturbation提供sameimagepair BradleyTerry监督，labelledvalidation只选layer。reward由sampler决定不被benchmark唯一识别；再用policy自身生成candidate与独立verifier标pairbootstraphead，和初始合成扰动训练区分。; 2B/8B/32B、MMRB1/VLRewardBench等，layerchoice比lastlayer+5.8points；相同bench分数的不同perturbationheads hallucination判断差6.4points，verifierbootstrap改善caption/imagegeneration但所有rewardbench下降。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "实施阶段已读摘要、作者信息、定义算法方法与主实验，以及列明的相关附录；复用实际实施审查，未宣称全部附录/整篇全文已读。", "review_artifact": "research-audits/oct12-review-closure/frozen-foundation-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12707v1", "sha256": "bf8fe4d9ecf041ecb1a8efb2caee64631527f1aa5627e740cfaeca358e8e6eb8" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "accepted", "implementation_key": "umrm", "accepted_scope": "bounded faithful mechanism/runtime or evaluation infrastructure; not original capability replication" }, { "id": "2610.12703", "title": "Bridging Human and Model Attention: Explainability Analysis of CNN, Mamba, and ViT Architectures with Gaze-Based Validation", "published": "2026-10-08T20:33:38Z", "reviews": [ { "id": "2610.12703", "title": "Bridging Human and Model Attention: Explainability Analysis of CNN, Mamba, and ViT Architectures with Gaze-Based Validation", "published": "2026-10-08T20:33:38Z", "decision": "reviewed_deferred", "reason": "pretrainedResNet/ViT/ViM任务adapt后各自XAI图与human-gaze heatmap统一normalized spatialprobabilities，比JSS/CSS/Pearson，分析freeze/ensemble。; fine-grained/medical/emotion三gazedata，Mamba出现near-unit supportoverlap却近零linearcorrelation。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12703v1#S4", "https://arxiv.org/html/2610.12703v1#S5", "https://arxiv.org/html/2610.12703v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12703v1", "sha256": "185652553eaef210de61b11e4a5d80e94b1538879af8858fdea14f4b7b49a5e0" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12668", "title": "Recurrent Harness for Agentic Continual Learning", "published": "2026-10-08T19:44:43Z", "reviews": [ { "id": "2610.12668", "title": "Recurrent Harness for Agentic Continual Learning", "published": "2026-10-08T19:44:43Z", "decision": "reviewed_deferred", "reason": "strict online forward-read检索过去反思，当前轨迹完成后逐个backward-revise被检索反思；修改只能源自原轨迹，当前轨迹仅反馈；revised reflection示范forward-write，无回放/validation搜索。; CL-Bench、MemoryBench、MemoryArena三backbone；source-grounding/泄漏审计、固定k=5后post-hoc敏感性、token和API成本。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12668v1#S3", "https://arxiv.org/html/2610.12668v1#S4", "https://arxiv.org/html/2610.12668v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12668v1", "sha256": "777a79b0dfea8cf1a8e9b70b30e9342cc059205ef6426572dad193c1c5775502" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12655", "title": "CLEAT: Learning Value-Guided Clarification for Interactive Agents", "published": "2026-10-08T19:22:21Z", "reviews": [ { "id": "2610.12655", "title": "CLEAT: Learning Value-Guided Clarification for Interactive Agents", "published": "2026-10-08T19:22:21Z", "decision": "reviewed_deferred", "reason": "已读CLEAT摘要、intent alignment及实验：同history ask/proceed paired rollout估residual query value，controller与task policy co-train，新policy outcome更新labels；Qwen3-4B四bench十设置三seed。下一批须训练真实value predictor+policy，不能拿测试latent goal作agent输入。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 Preliminaries", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12655v1", "sha256": "56f9af828d7d513b96e62c65177910b949fdcfcf16da02d4de78eb80d42498dd", "local_html_sha256": "56f9af828d7d513b96e62c65177910b949fdcfcf16da02d4de78eb80d42498dd", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12655v1", "result": "retrieved", "sha256": "56f9af828d7d513b96e62c65177910b949fdcfcf16da02d4de78eb80d42498dd" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12651", "title": "SynAct: Synergistic Active Feedback Learning for LLM Agents", "published": "2026-10-08T19:20:10Z", "reviews": [ { "id": "2610.12651", "title": "SynAct: Synergistic Active Feedback Learning for LLM Agents", "published": "2026-10-08T19:20:10Z", "decision": "reviewed_deferred", "reason": "已读SynAct方法/实验：重复rollout中action-terminal outcome互信息选择高价值decision问expert，distill成可复用rule，disjoint训练/测试；outcome恒零success时仍可按失败状态区分。下一批需真实反事实样本与预算matched routing，不是额外无差别retry。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 SynAct: Synergistic Active Feedback Learning", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12651v1", "sha256": "a9812088b48443f26f9627bf2e3b37b8599b55acbf6492f47a9adeca0f2b9e02", "local_html_sha256": "a9812088b48443f26f9627bf2e3b37b8599b55acbf6492f47a9adeca0f2b9e02", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12651v1", "result": "retrieved", "sha256": "a9812088b48443f26f9627bf2e3b37b8599b55acbf6492f47a9adeca0f2b9e02" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12650", "title": "Feature Competition and Its Representational Origins in Vision Language Models: Evidence from the Stroop Paradigm", "published": "2026-10-08T19:13:43Z", "reviews": [ { "id": "2610.12650", "title": "Feature Competition and Its Representational Origins in Vision Language Models: Evidence from the Stroop Paradigm", "published": "2026-10-08T19:13:43Z", "decision": "reviewed_deferred", "reason": "Stroop100 word×ink stimuli比较contrastive/generativeVLM行为，ink-only/word-only/combinedembedding RSA与conceptchunk centroidsteering分离文字/颜色feature影响。; 10congruent90incongruent，embedintervention仅CLIP；wordsteering较易、inkdirection弱且generative输出分类有edgecases。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12650v1#Sx3", "https://arxiv.org/html/2610.12650v1#Sx4", "https://arxiv.org/html/2610.12650v1#Sx5", "https://arxiv.org/html/2610.12650v1#Sx8" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12650v1", "sha256": "1e375ee521f2d59ae9e54495bf5ce1f8bfa11756fa0b653a94a8677d8df75101" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12611", "title": "SoftwareGym: Scaling Computer-Use Agent Training with Diverse Synthesized Software Environments", "published": "2026-10-08T18:31:16Z", "reviews": [ { "id": "2610.12611", "title": "SoftwareGym: Scaling Computer-Use Agent Training with Diverse Synthesized Software Environments", "published": "2026-10-08T18:31:16Z", "decision": "reviewed_deferred", "reason": "已读SoftwareGym方法/实验：合成真实可执行GUI app、固定v1 GUI测试迭代repair，182app12类120task，Qwen3.5-9B multimodal SFT、out-of-domain OSWorld。下一批须VM实际GUI交互与合成环境自测隔离，不能用静态截图/固定动作fixture。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12611v1", "sha256": "eb033a8b472d058f4a31e948e0254623f6781140b5131439fca652338d66350c", "local_html_sha256": "eb033a8b472d058f4a31e948e0254623f6781140b5131439fca652338d66350c", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12611v1", "result": "retrieved", "sha256": "eb033a8b472d058f4a31e948e0254623f6781140b5131439fca652338d66350c" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12602", "title": "Self-Preservation in Language Models", "published": "2026-10-08T18:24:02Z", "reviews": [ { "id": "2610.12602", "title": "Self-Preservation in Language Models", "published": "2026-10-08T18:24:02Z", "decision": "reviewed_deferred", "reason": "Task Bank中的selector、frozen pre-update replay generator、learner分别执行；只混合incoming与自生成replay，held-out任务描述/答案/分数不进入选bank。; §3含Qwen0.6/1.7/4B多seed公开任务、matched pre/post retention；4B incoming-aware selection优于random，但小模型无对应收益。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract/author information/main defining methods/main experiments and tables; production-evidence paragraph spot check; appendices not exhaustively read", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12602v1#S2", "https://arxiv.org/html/2610.12602v1#S3" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12602v1", "sha256": "2e2ca79ebe6e8e0b0cbbcecd8fd2a8113dcbdfceaf604845e9b151371c08badf" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12601", "title": "Reflective Prompt Optimization of a Deep-Research Agent: A Topic-Disjoint, Multi-Seed Protocol for Measuring Transferable Gains and Prompt-Level Overfitting", "published": "2026-10-08T18:23:23Z", "reviews": [ { "id": "2610.12601", "title": "Reflective Prompt Optimization of a Deep-Research Agent: A Topic-Disjoint, Multi-Seed Protocol for Measuring Transferable Gains and Prompt-Level Overfitting", "published": "2026-10-08T18:23:23Z", "decision": "reviewed_deferred", "reason": "GEPA提示进化的topic-disjoint 500-task评测和verifier ablation可用于评测完整性；属于冻结模型prompt/agent优化，下一批转agent/evaluation，并需隔离winning prompt内answer-key风险。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12601v1", "sha256": "b639966ea3e1b51667cf8817044471892cfb515025d18323caa410f22d74cd69", "local_html_sha256": "b639966ea3e1b51667cf8817044471892cfb515025d18323caa410f22d74cd69", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12601v1", "result": "retrieved", "sha256": "b639966ea3e1b51667cf8817044471892cfb515025d18323caa410f22d74cd69" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.12598", "title": "Beyond Binary Verifiable Rewards: Behavioral Self-Rewarding for SWE Agents", "published": "2026-10-08T18:21:10Z", "reviews": [ { "id": "2610.12598", "title": "Beyond Binary Verifiable Rewards: Behavioral Self-Rewarding for SWE Agents", "published": "2026-10-08T18:21:10Z", "decision": "reviewed_deferred", "reason": "已读BSR方法/实验：同task同verifier bucket内policy比较debugging行为，bounded辅助reward保留pass/fail序；post-rollout训练judge显式看训练reference patch，部署/评测agent不得读gold。下一批必须真实SWE RL更新，不能仅给轨迹打heuristic分。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12598v1", "sha256": "4204fe00e9ec30d63a19deab3001bb36bb13c02aeebb323d64126d3e859b707a", "local_html_sha256": "4204fe00e9ec30d63a19deab3001bb36bb13c02aeebb323d64126d3e859b707a", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12598v1", "result": "retrieved", "sha256": "4204fe00e9ec30d63a19deab3001bb36bb13c02aeebb323d64126d3e859b707a" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12596", "title": "SpireBench: Evaluating Long-Horizon Strategic Play of Agentic Models with Slay the Spire", "published": "2026-10-08T18:18:19Z", "reviews": [ { "id": "2610.12596", "title": "SpireBench: Evaluating Long-Horizon Strategic Play of Agentic Models with Slay the Spire", "published": "2026-10-08T18:18:19Z", "decision": "reviewed_deferred", "reason": "已读SpireBench方法/结果：完整commercial Slay the Spire长程决策、全术语reskin对prior知识、跨run memory控制与cost frontier；游戏license/真实runtime待核，下一批benchmark infra，不能象征性替代完整游戏。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Methods", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12596v1", "sha256": "109490a1dae512c8b7e5486fab1261a86811ca55f44528185a93d6c15df60ace", "local_html_sha256": "109490a1dae512c8b7e5486fab1261a86811ca55f44528185a93d6c15df60ace", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12596v1", "result": "retrieved", "sha256": "109490a1dae512c8b7e5486fab1261a86811ca55f44528185a93d6c15df60ace" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12593", "title": "Know What You Don't Know: Agentic Data Synthesis by Evolving Generation Space for Closed-Loop Learning", "published": "2026-10-08T18:17:20Z", "reviews": [ { "id": "2610.12593", "title": "Know What You Don't Know: Agentic Data Synthesis by Evolving Generation Space for Closed-Loop Learning", "published": "2026-10-08T18:17:20Z", "decision": "reviewed_deferred", "reason": "已读EvoSyn方法/实验：可修订synthesis state将learner弱点转成within-space分配与新factor扩空间，Designer/Planner生成真实环境经验再训练learner，tau2/CAR benchmark。下一批需保留data synthesis–training闭环，固定生成题库不等同算法。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12593v1", "sha256": "5fa211065130631e28ce1e64f3fe01d4b64a64863ffcd979c446829acfc0cf10", "local_html_sha256": "5fa211065130631e28ce1e64f3fe01d4b64a64863ffcd979c446829acfc0cf10", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12593v1", "result": "retrieved", "sha256": "5fa211065130631e28ce1e64f3fe01d4b64a64863ffcd979c446829acfc0cf10" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12582", "title": "One Model to Loop Them All: Elastic Latent Recurrence for Language Models", "published": "2026-10-08T18:05:42Z", "reviews": [ { "id": "2610.12582", "title": "One Model to Loop Them All: Elastic Latent Recurrence for Language Models", "published": "2026-10-08T18:05:42Z", "decision": "reviewed_deferred", "reason": "ELR实际共享latent block、learned gated step、loop/budget conditioning与每update mixed-depth训练；4B dense/30B-A3B评测及route reuse/KV tradeoff，foundation recurrence下一批，不能只插循环标签冒充模型训练。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 One-Loop Accuracy and Gains from Additional Loops", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12582v1", "sha256": "299b298ab2cc380e7ee9c422e1f7d74ac3823b7643a84481ec681ff41fb87d16", "local_html_sha256": "299b298ab2cc380e7ee9c422e1f7d74ac3823b7643a84481ec681ff41fb87d16", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12582v1", "result": "retrieved", "sha256": "299b298ab2cc380e7ee9c422e1f7d74ac3823b7643a84481ec681ff41fb87d16" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.12575", "title": "Mind the Temporal Gap: Evaluating Video Models for Temporal Understanding in Driving", "published": "2026-10-08T18:02:22Z", "reviews": [ { "id": "2610.12575", "title": "Mind the Temporal Gap: Evaluating Video Models for Temporal Understanding in Driving", "published": "2026-10-08T18:02:22Z", "decision": "reviewed_deferred", "reason": "egomotion multi-levelprobe比较standard/visual/fewshot/CoT，LoRA和视觉特征压缩前/后1layerCNN probe定位信息缺失或语言readout失败。; CARLA controlledsim与小realdriving仅standardprompt，较新模型fewshot有taskgain但类biaspersist；learning/probe结果限Sim。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12575v1#S2", "https://arxiv.org/html/2610.12575v1#S3", "https://arxiv.org/html/2610.12575v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12575v1", "sha256": "24cbdd35b322d185a135019a40ad473f0f119f204f4ff319838dacbea0cb7285" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12569", "title": "SWE-Duel: Towards a Self-Scaling Arena for Coding Agents", "published": "2026-10-08T18:01:00Z", "reviews": [ { "id": "2610.12569", "title": "SWE-Duel: Towards a Self-Scaling Arena for Coding Agents", "published": "2026-10-08T18:01:00Z", "decision": "reviewed_deferred", "reason": "已读SWE-Duel arena与实验：Red context隔离造feature+sealed bugtests，admission含完整test一致性与fresh blind repair，Blue不知道bug；1500turn11participant15repo。下一批真实任务生成/repair，封存oracle不可被Blue读取，BT排名非静态gold。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 SWE-Duel : A Self-scaling Coding Arena", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12569v1", "sha256": "17efd0bf9d7f1f02f32113cf8c51d83f1930b473269a4b086123b69bdc2f4c54", "local_html_sha256": "17efd0bf9d7f1f02f32113cf8c51d83f1930b473269a4b086123b69bdc2f4c54", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12569v1", "result": "retrieved", "sha256": "17efd0bf9d7f1f02f32113cf8c51d83f1930b473269a4b086123b69bdc2f4c54" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12566", "title": "Sample-Adaptive Zeroth-Order Post-Training", "published": "2026-10-08T18:00:37Z", "reviews": [ { "id": "2610.12566", "title": "Sample-Adaptive Zeroth-Order Post-Training", "published": "2026-10-08T18:00:37Z", "decision": "reviewed_deferred", "reason": "官方PDF已取得并限定阅读算法及评测。SAZO复用two-forward directional difference，按per-example EMA估计梯度范数重要性，带uniform floor importance sampling及1/(Np_i)无偏缩放，仅更新抽中样本score；原文有RoBERTa/OPT/Phi/Llama matched-budget实证。冻结16篇之外，下一批需真实ZO参数扰动/恢复、随机方向与采样账本，不能常量loss或固定重权代理。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_official_pdf_method_and_evaluation_pages", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "Title/author and arXiv v1 margin", "pdf_page": 1, "scope": "full page text read; manuscript cover date is not substituted for arXiv v1 date" }, { "heading": "Method and evaluation/conclusion", "pdf_pages": [ 3, 4, 6, 7, 8, 9 ], "scope": "listed page text read; not all article pages or appendices", "method_evidence": "s_i=(f(x+mu z)-f(x-mu z))/(2mu)；p_i=(1-eps)m_i/sum(m)+eps/N；update=s_i z/(N p_i)；仅selected m_i做(1-beta)m_i+beta sqrt(pi/2)|s_i|，其他score不变；无额外ZO查询、O(N)scalar state。", "evaluation_evidence": "§4/Table2: RoBERTa-large350M六任务full/PEFT；Table3: OPT1.3B/Phi2/Llama3-8B八任务并同模型LR search grid；Figure2在4k/6k/6k forward passes达到MeZO24k端点，不能直接称wallclock同倍加速；Table4 latency与MeZO相近；population scaling三seed；beta1发散，.005/.05/.5稳定且.05最佳。" } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12566v1", "sha256": null, "local_html_sha256": null, "local_html_sha256_matches": null, "attempts": [ { "url": "https://arxiv.org/html/2610.12566v1", "result": "failed", "error": "404 Client Error: Not Found for url: https://arxiv.org/html/2610.12566v1" }, { "url": "https://arxiv.org/html/2610.12566v1", "result": "failed", "error": "404 Client Error: Not Found for url: https://arxiv.org/html/2610.12566v1" }, { "url": "https://arxiv.org/pdf/2610.12566v1", "result": "retrieved", "retrieved_on": "2026-10-12", "command": "curl -sSL --fail --max-time 45 https://arxiv.org/pdf/2610.12566v1 -o tmp/pdfs/2610.12566.pdf", "sha256": "109824aa290f19c36607545f5579a3472def7a2bf9c5793f0b13c31aa5a4b2f2" } ], "alternate_evidence": { "url": "https://arxiv.org/pdf/2610.12566v1", "format": "official_arxiv_pdf", "sha256": "109824aa290f19c36607545f5579a3472def7a2bf9c5793f0b13c31aa5a4b2f2", "local_path": "tmp/pdfs/2610.12566.pdf", "local_sha256_verified": true, "total_pdf_pages": 23 } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.12545", "title": "AutoIndexer: Flexible-Order Decoding by Training Causal Models on Chains of Edits", "published": "2026-10-08T17:59:52Z", "reviews": [ { "id": "2610.12545", "title": "AutoIndexer: Flexible-Order Decoding by Training Causal Models on Chains of Edits", "published": "2026-10-08T17:59:52Z", "decision": "reviewed_deferred", "reason": "AutoIndexer用inverse chains of edits训练causal model cursor/insert/delete/substitute，Qwen3-8B LoRA extra100k CPT与原AR/LLM benchmarks；下一批真实cursor与edit生成语义，不能后处理edit heuristic冒充模型。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "2 AutoIndexer", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "3 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12545v1", "sha256": "5e0bec696fb8d5d4ebc4c72d8cd44acc5c5ff77f4f4f2fa4b1750081772ed5ff", "local_html_sha256": "5e0bec696fb8d5d4ebc4c72d8cd44acc5c5ff77f4f4f2fa4b1750081772ed5ff", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12545v1", "result": "retrieved", "sha256": "5e0bec696fb8d5d4ebc4c72d8cd44acc5c5ff77f4f4f2fa4b1750081772ed5ff" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.12461", "title": "OuroWorld: Bringing Any 3D World Alive as Diverse, Endlessly Looping 3D Cinemagraphs", "published": "2026-10-08T17:59:46Z", "reviews": [ { "id": "2610.12461", "title": "OuroWorld: Bringing Any 3D World Alive as Diverse, Endlessly Looping 3D Cinemagraphs", "published": "2026-10-08T17:59:46Z", "decision": "reviewed_deferred", "reason": "OuroWorld用VLM选referenceview/推dynamic→真实videogeneration，再VGGTlift/TrajectoryCrafter补多view；Fourierperiodic4DGS deformation与reference-groundedDriftField吸收viewinconsistency。; 39 reconstructed/generated3DGS scenes，GTfreeautomaticmetrics+humanpairwise；去periodiccycle seam坏而overallmotion指标不明显变。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12461v1#S3", "https://arxiv.org/html/2610.12461v1#S4", "https://arxiv.org/html/2610.12461v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12461v1", "sha256": "1e697597bea79be0b8a3803246d215c0651dfdf482e375ad5796339b77af9d61" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12458", "title": "OmniCapBench: A Deep-Structured Evaluation Framework for Fine-Grained Audio-Visual Captioning", "published": "2026-10-08T17:59:21Z", "reviews": [ { "id": "2610.12458", "title": "OmniCapBench: A Deep-Structured Evaluation Framework for Fine-Grained Audio-Visual Captioning", "published": "2026-10-08T17:59:21Z", "decision": "reviewed_deferred", "reason": "OmniCapBench model直接从真实audio-video生成entityreference/shot/audioevent三trackatomicJSONunits；先deterministicstructural约束再localizedsemanticLLMcompare，避免wholecaption单分数。; 786densevideos多omnimodel；specializedfreeformcaptioners因structuredinstructionpoor被排除主表，identitydrift/crossmodalalignment是难点。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12458v1#S3", "https://arxiv.org/html/2610.12458v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12458v1", "sha256": "d2ad8894e91f5707e79165f43f645125538db971e99b9fbca5904c623d911936" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12451", "title": "VersaCamVLA: Camera-Configurable VLA Policies for Robotic Manipulation", "published": "2026-10-08T17:58:47Z", "reviews": [ { "id": "2610.12451", "title": "VersaCamVLA: Camera-Configurable VLA Policies for Robotic Manipulation", "published": "2026-10-08T17:58:47Z", "decision": "reviewed_deferred", "reason": "VersaCamVLA把variableposedRGB集合成fixedscenetokens，stageI RGB/semantic/edge targetview预测+WristAugmentedPoseSampling；stageII丢decoder、frozenceneencoder+jointspatialcompressor补pretrainedVLA视觉条件。; RoboTwin2/LIBERO和realrobot camera-count/unseenpose，combinesemantic+edge比RGBonly更好。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12451v1#S3", "https://arxiv.org/html/2610.12451v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12451v1", "sha256": "3ccedc1057a0bb84732928d9f4d373d522e7e3962c42350685727d06b2e988a2" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12445", "title": "Caught in the Act: Probes Effectively Detect Sabotage and Catch Unverbalized Deception", "published": "2026-10-08T17:58:13Z", "reviews": [ { "id": "2610.12445", "title": "Caught in the Act: Probes Effectively Detect Sabotage and Catch Unverbalized Deception", "published": "2026-10-08T17:58:13Z", "decision": "reviewed_deferred", "reason": "已读FIBS/probe方法及结果：多layer/token白盒probe，163016上下文falsehood数据，SHADE-Arena honest heldout校准以及introspective deception独立elicitation。下一批需真实hidden activations训练、场景分割和FPR校准，不把AUC99当安全保证。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12445v1", "sha256": "cda4538c8e0901e4eb02030b4ece8c77e86ef40460389168feca9273de834d36", "local_html_sha256": "cda4538c8e0901e4eb02030b4ece8c77e86ef40460389168feca9273de834d36", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12445v1", "result": "retrieved", "sha256": "cda4538c8e0901e4eb02030b4ece8c77e86ef40460389168feca9273de834d36" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12427", "title": "FastBench: Can Streaming VLMs Perceive High-Dynamic Real-World Streams?", "published": "2026-10-08T17:55:20Z", "reviews": [ { "id": "2610.12427", "title": "FastBench: Can Streaming VLMs Perceive High-Dynamic Real-World Streams?", "published": "2026-10-08T17:55:20Z", "decision": "reviewed_deferred", "reason": "FastBench高FPStrajectory-groundedQA过滤2FPS可答样本、人核evidenceinterval；ProactiveFrame在1secstreamchunk通过texttoken调整未来FPS，近期高FPSFocusWindow+较旧sparseGeneralWindow。; 306QA、forward/instant/backward三时序scope与多domain，denserframes24FPS后饱和/30FPS下降；oracleintervalfocusing明显强于model自主rate。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12427v1#S3", "https://arxiv.org/html/2610.12427v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12427v1", "sha256": "f561945a7ffb1fc8cf52dffb5cd54ee89098999411b1574be0f2dd675db50887" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12420", "title": "A Unified Bellman Operator for Safety-Critical Reinforcement Learning", "published": "2026-10-08T17:53:02Z", "reviews": [ { "id": "2610.12420", "title": "A Unified Bellman Operator for Safety-Critical Reinforcement Learning", "published": "2026-10-08T17:53:02Z", "decision": "rejected_or_out_of_scope", "reason": "统一Bellman safety/task value与两时间尺度收敛主要面向neural continuous control；未涉及LLM/foundation/agent语言后训练，当前研究批次out of scope，不因强化学习关键词纳入。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Learning Safety-Constrained Task Action-Value Function", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "7 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12420v1", "sha256": "7beeff8ccaa4d70298d7ed5f54eba7a7bc07043a4f08f0a0c99272f11ac0c8cd", "local_html_sha256": "7beeff8ccaa4d70298d7ed5f54eba7a7bc07043a4f08f0a0c99272f11ac0c8cd", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12420v1", "result": "retrieved", "sha256": "7beeff8ccaa4d70298d7ed5f54eba7a7bc07043a4f08f0a0c99272f11ac0c8cd" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.12417", "title": "WOVEN: Weaving Visual World Modeling into Multimodal LLMs", "published": "2026-10-08T17:52:03Z", "reviews": [ { "id": "2610.12417", "title": "WOVEN: Weaving Visual World Modeling into Multimodal LLMs", "published": "2026-10-08T17:52:03Z", "decision": "accepted", "reason": "真实transition图像条件监督，source/action/reasoning三轴、真实VLMresponse-only SFT和optiontargetpermutation。; 已实现publicWOVENtrain mini-suite与真实VLM链路，source_id分组隔离；作者officialvalidation/test未发布，GPU未完成。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract/author information/main defining methods/main experiments and tables; production-evidence paragraph spot check; appendices not exhaustively read", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12417v1#S2", "https://arxiv.org/html/2610.12417v1#S3", "https://arxiv.org/html/2610.12417v1#A7.SS1" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12417v1", "sha256": "d6a25232b04ca31e9ca7f1b2c30de215a8ed8e4317dabe1b7044130866676011" }, "new_implementation_inferred_from_review": false }, { "id": "2610.12417", "title": "WOVEN: Weaving Visual World Modeling into Multimodal LLMs", "published": "2026-10-08T17:52:03Z", "decision": "accepted", "reason": "以真实视觉transitionrollout、多scene/action/reasoningtype生成supervision，给真实MLLMvisualinput与question/transitiontarget进行SFT/LoRA；通过reasoningoperation而非scene/action/domain选择训练supervision、statechange幅度检验robustness。不是静态imageclassscore替代visualworldmodel。; 原文36076examples/20scene/5action/8reasoning、38MLLM及human，约2000item不同subsets训练，26externalbench中的22有改善、taskdatareplacement与heldoutrecipeprospectivevalidation。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "实施阶段已读摘要、作者信息、定义算法方法与主实验，以及列明的相关附录；复用实际实施审查，未宣称全部附录/整篇全文已读。", "review_artifact": "research-audits/oct12-review-closure/frozen-foundation-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12417v1", "sha256": "d6a25232b04ca31e9ca7f1b2c30de215a8ed8e4317dabe1b7044130866676011" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "accepted", "implementation_key": "woven", "accepted_scope": "bounded faithful mechanism/runtime or evaluation infrastructure; not original capability replication" }, { "id": "2610.12410", "title": "Predicting Alignment Generalization with Value Representations", "published": "2026-10-08T17:47:26Z", "reviews": [ { "id": "2610.12410", "title": "Predicting Alignment Generalization with Value Representations", "published": "2026-10-08T17:47:26Z", "decision": "reviewed_deferred", "reason": "Value representations通过single-value DPO/activation spaces预测66-value alignment generalization，并研究multivalue robustness；下一批需真实finetune与heldout value matrix，predictive correlations不能当causal controls保证。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Computing Alignment Generalization Effects", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Which Value Representations Best Predict Alignment Generalization?", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12410v1", "sha256": "86b49a8ca953b909c9ee9eaeb0e3450be53276ebb8a9973dcc0f10cb2d9d3c9a", "local_html_sha256": "86b49a8ca953b909c9ee9eaeb0e3450be53276ebb8a9973dcc0f10cb2d9d3c9a", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12410v1", "result": "retrieved", "sha256": "86b49a8ca953b909c9ee9eaeb0e3450be53276ebb8a9973dcc0f10cb2d9d3c9a" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.12402", "title": "SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models", "published": "2026-10-08T17:44:17Z", "reviews": [ { "id": "2610.12402", "title": "SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models", "published": "2026-10-08T17:44:17Z", "decision": "reviewed_deferred", "reason": "SpaceCast observe-transform-infer只提供变换前orderedRGBviews，自然语言T之后模型推未观察S1；programmaticverification分static/local/global预测及桥接views控制。; 182realScanNet/+scenes3862Q、21models，syntheticSFT/GRPO和sixOOD；explicit3D evidence比生成outcome媒介更稳。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12402v1#S3", "https://arxiv.org/html/2610.12402v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12402v1", "sha256": "9073196d483567b87899fd3b6a0195860683a1f4404fdcadb8da71cccbfbbefc" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12390", "title": "Long Text to Predictive Features: LLM-Guided Blockwise Feature Engineering via Executable Program Search", "published": "2026-10-08T17:39:11Z", "reviews": [ { "id": "2610.12390", "title": "Long Text to Predictive Features: LLM-Guided Blockwise Feature Engineering via Executable Program Search", "published": "2026-10-08T17:39:11Z", "decision": "reviewed_deferred", "reason": "LLMBlockFE offline只appendimmutablefeaturecodeblock，真实execute/downstreamAUCgate+localBayesianparamtuning；depthcalibratedcreditrollback，多independenttrajectory只共享fixed方向描述，freezebestprograms供online不调用LLM。; CFPBTimely/GossipCop+两private，五生产riskapps2025.9–11同样本parallelKS比较2694342用户；原文另报告onlineAB约RMB1B放款、审批数不变、baddebt8.0%→7.7%。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12390v1#S3", "https://arxiv.org/html/2610.12390v1#S4", "https://arxiv.org/html/2610.12390v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12390v1", "sha256": "6d8a45d94d93f77faf6a0516a886047cd2d50497d47e61bc5f422dc56e7be372" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12382", "title": "WorldAlign: Decoupled 4D Reward for World-Consistent Video Generation", "published": "2026-10-08T17:35:19Z", "reviews": [ { "id": "2610.12382", "title": "WorldAlign: Decoupled 4D Reward for World-Consistent Video Generation", "published": "2026-10-08T17:35:19Z", "decision": "reviewed_deferred", "reason": "语义subjectmasktracker而非motion检测分静/动；GFM staticpointcloud maskedreprojection+camera-motion hinge防staticcollapse；frozenVLM sample-specificdynamic/physics/shape/texturebinarychecklistreward，samepromptgroup各rewardnormalize后GRPO。; DAVIS/COCO1389samples，100randomheldout；Wan2.1/Wan2.2 matcheddata/settings，rewardindependentmaskedepipolar及trackedDINO/vbench检查。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "official PDF pages1..7 first3000 extracted characters each; not complete15pages", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/pdf/2610.12382v1#page=5", "https://arxiv.org/pdf/2610.12382v1#page=6", "https://arxiv.org/pdf/2610.12382v1#page=7" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/pdf/2610.12382v1", "sha256": "32a7dc1682308baad3e18d3653c10fc11c15ff7d2c7fabbbdd67472970c4b479" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12375", "title": "OnTrack: Real-Time Monitoring and Intervention in LLM Agent Trajectories via Streaming Structure-Aware Optimal Transport", "published": "2026-10-08T17:32:08Z", "reviews": [ { "id": "2610.12375", "title": "OnTrack: Real-Time Monitoring and Intervention in LLM Agent Trajectories via Streaming Structure-Aware Optimal Transport", "published": "2026-10-08T17:32:08Z", "decision": "reviewed_deferred", "reason": "已读OnTrack问题/实验：stream dependency graph的结构optimal transport，full/intermediate/zero prior三regime；synthetic unit embedding只是fault diagnostic，2294真实SWE trajectory streaming replay另记。下一批必须防成功reference泄漏测试任务，并区分offline replay与在线block权限。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "6 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Problem Formulation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12375v1", "sha256": "6f38f40ae058e74bdee9d5dbb3f6a50df9951b543494bffe2e79feeea6cb4141", "local_html_sha256": "6f38f40ae058e74bdee9d5dbb3f6a50df9951b543494bffe2e79feeea6cb4141", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12375v1", "result": "retrieved", "sha256": "6f38f40ae058e74bdee9d5dbb3f6a50df9951b543494bffe2e79feeea6cb4141" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12367", "title": "Which Skill to Distill? SGUID: Selecting a Compact Skill Bank for Model-Skill Co-Evolution", "published": "2026-10-08T17:27:53Z", "reviews": [ { "id": "2610.12367", "title": "Which Skill to Distill? SGUID: Selecting a Compact Skill Bank for Model-Skill Co-Evolution", "published": "2026-10-08T17:27:53Z", "decision": "accepted", "reason": "SGUID已implemented：按skill-conditioned teachers对student trajectory的有效learning signals选compact bank，并updated-model rounds重建bank；DAPO/math benchmark与bank controls，保留existing实现状态，不宣称本批新实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12367v1", "sha256": "44936fbd5a055ab8961f343749b98311cbe44d4b1f79d75d367cc23b3757a0c7", "local_html_sha256": "44936fbd5a055ab8961f343749b98311cbe44d4b1f79d75d367cc23b3757a0c7", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12367v1", "result": "retrieved", "sha256": "44936fbd5a055ab8961f343749b98311cbe44d4b1f79d75d367cc23b3757a0c7" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "accepted" }, { "id": "2610.12360", "title": "Accurate but Not Humble: Evaluating Epistemic Humility in LLM Agents under Knowledge Conflict", "published": "2026-10-08T17:25:06Z", "reviews": [ { "id": "2610.12360", "title": "Accurate but Not Humble: Evaluating Epistemic Humility in LLM Agents under Knowledge Conflict", "published": "2026-10-08T17:25:06Z", "decision": "reviewed_deferred", "reason": "已读EH conflict构建和findings：controlled/natural conflict配matched controls，trajectory Identify/Solve/Escalate，任务accuracy与显式uncertainty分开。下一批适合行为eval，gold belief elicitation仅task构造/外部评分，不灌被评agent。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 Evaluating Epistemic Humility via Knowledge Conflict", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Findings", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12360v1", "sha256": "86642e62300a268ab94004236bb61785247e32711b2d0f3e244dd06c29239f9d", "local_html_sha256": "86642e62300a268ab94004236bb61785247e32711b2d0f3e244dd06c29239f9d", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12360v1", "result": "retrieved", "sha256": "86642e62300a268ab94004236bb61785247e32711b2d0f3e244dd06c29239f9d" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12355", "title": "Distilling Routed 3D Privilege for Spatial Reasoning in Vision-Language Models", "published": "2026-10-08T17:22:24Z", "reviews": [ { "id": "2610.12355", "title": "Distilling Routed 3D Privilege for Spatial Reasoning in Vision-Language Models", "published": "2026-10-08T17:22:24Z", "decision": "reviewed_deferred", "reason": "GPD offline从3Dscan生成depth/semantic/BEV compacttext且questionconditionroute给同模型teacher+trainreferenceanswer；GRPO加privilegedKL仅incorrectstudenttrajectories，deployRGBonly。; 五spatialbench samebackbone/dataGRPO/OPSD controlled对照，3D+answerprivilege互补、只wrongKL与routing消融。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12355v1#S3", "https://arxiv.org/html/2610.12355v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12355v1", "sha256": "c3e7a85930bb55c1d3a373b2488d2ebb323227856a8db456fb1874acdd6e3153" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12341", "title": "Can AI Agents Learn Their Way to the Top? Evaluating Heuristic Learning in a Long-Running Game Agent Competition", "published": "2026-10-08T17:15:48Z", "reviews": [ { "id": "2610.12341", "title": "Can AI Agents Learn Their Way to the Top? Evaluating Heuristic Learning in a Long-Running Game Agent Competition", "published": "2026-10-08T17:15:48Z", "decision": "reviewed_deferred", "reason": "AHL在固定模型权重下修改可执行game policies，AAArena人类ladder预算评测；主要是agent程序优化而非参数后训练，转agent track下一批，不能登记LLM权重训练复现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "2 Adversarial Heuristic Learning", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12341v1", "sha256": "b41615e1af7afaeb2188eca093e6b300a43e3c83f393407abbd5d3b82f7c438e", "local_html_sha256": "b41615e1af7afaeb2188eca093e6b300a43e3c83f393407abbd5d3b82f7c438e", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12341v1", "result": "retrieved", "sha256": "b41615e1af7afaeb2188eca093e6b300a43e3c83f393407abbd5d3b82f7c438e" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.12338", "title": "VFold: Symmetry-Aware Cross-Layer Value Cache Compression", "published": "2026-10-08T17:14:28Z", "reviews": [ { "id": "2610.12338", "title": "VFold: Symmetry-Aware Cross-Layer Value Cache Compression", "published": "2026-10-08T17:14:28Z", "decision": "reviewed_deferred", "reason": "VFold CCAcalibration得adjacentvaluebasisalignment并fold WV/WO functionpreservingweights，runtimevalues用后merge到sharedrunningmean跨layers；keys/RoPE保持不合并。; Llama8B/Mistral24B/Qwen8B LongBench/RULER，50%value减仅25%totalKV；还有quant/keyprune组合。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12338v1#S3", "https://arxiv.org/html/2610.12338v1#S4", "https://arxiv.org/html/2610.12338v1#S5", "https://arxiv.org/html/2610.12338v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12338v1", "sha256": "8f2f7694c5a6f8d55d59d33e75d21092cdbcf5eaf87fdaa0809ce694d7e2cc8d" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12327", "title": "SparseDecoding: Decoding-Aware Pruning for Accurate and Efficient LLM Inference", "published": "2026-10-08T17:06:32Z", "reviews": [ { "id": "2610.12327", "title": "SparseDecoding: Decoding-Aware Pruning for Accurate and Efficient LLM Inference", "published": "2026-10-08T17:06:32Z", "decision": "reviewed_deferred", "reason": "SparseDecoding真实dense ARgeneration收layeractivations且排除prefill形成Hessian/calibration矩阵，再OBS/SparseGPT/Wanda剪枝；N:MSpMV custombitmask索引fixed-step遍历针对decodebatch1。; Llama8/70B与Qwen14/32B长文/codegen质量，A100nativeSpMV end-to-enddecode测量与C4cal对照。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12327v1#S3", "https://arxiv.org/html/2610.12327v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12327v1", "sha256": "7a670983a55e4b6bc7f4c2843a586a59773b4365edef7545d72b369053a536b8" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12325", "title": "Prior or Feedback? What an LLM Uses When Adapting Neural Operators", "published": "2026-10-08T17:06:23Z", "reviews": [ { "id": "2610.12325", "title": "Prior or Feedback? What an LLM Uses When Adapting Neural Operators", "published": "2026-10-08T17:06:23Z", "decision": "reviewed_deferred", "reason": "已读budgeted FNO adaptation及controlled intervention：20trial matched预算，LLM根据任务语义和validation history选finetune参数，score置换/保值rewrite分别验证反馈敏感性。heldout nRMSE仅最终评测，下一批autonomous research机制不以固定fitness代理真实训练。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 Experimental setting: budgeted FNO adaptation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Controlled input interventions", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12325v1", "sha256": "f31b4c746eb52ef06f9a94fc8e055e56e326776f104aab90d856c931ffdf17dd", "local_html_sha256": "f31b4c746eb52ef06f9a94fc8e055e56e326776f104aab90d856c931ffdf17dd", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12325v1", "result": "retrieved", "sha256": "f31b4c746eb52ef06f9a94fc8e055e56e326776f104aab90d856c931ffdf17dd" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12310", "title": "Is In-Domain Training Enough for Fine-Grained Industrial Anomaly Understanding?", "published": "2026-10-08T16:59:29Z", "reviews": [ { "id": "2610.12310", "title": "Is In-Domain Training Enough for Fine-Grained Industrial Anomaly Understanding?", "published": "2026-10-08T16:59:29Z", "decision": "reviewed_deferred", "reason": "SiGMA normalreference/industrialknowledgesearch+heterogeneousMLLMreaders+真实visualdefectexpert，query-ref evidencecalibration；unlabeledpredictions估taskcompetence/queryevidence质量后reliability聚合。; MMAD七task，trainedspecialist定位仍弱/detection可差于base；agent-only互补不能补fineperception，expert reader有call/cost增加。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12310v1#S3", "https://arxiv.org/html/2610.12310v1#S4", "https://arxiv.org/html/2610.12310v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12310v1", "sha256": "988229c4188a39ba3e189f84a70f929aebc6fa1c94f2953c0c42699a3420ea15" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12305", "title": "From What to Which: Decoding Modifier Grounding in Frozen MLLMs", "published": "2026-10-08T16:53:59Z", "reviews": [ { "id": "2610.12305", "title": "From What to Which: Decoding Modifier Grounding in Frozen MLLMs", "published": "2026-10-08T16:53:59Z", "decision": "reviewed_deferred", "reason": "OTTER frozenMLLMgenerated-token/visualpatchstates监督投影，semantic/attention/spatialcost和learneddustbin/marginals做entropyregularizedSinkhornOT，compactpatchmap+optionalresidual。; RefCOCO controlledmodifier/noun、perturb/freegeneration/RefCOCOgtransfer；transfer用fixedfullOT而主表adaptivegated，backbonelayerselection另appendix。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12305v1#S3", "https://arxiv.org/html/2610.12305v1#S4", "https://arxiv.org/html/2610.12305v1#S5", "https://arxiv.org/html/2610.12305v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12305v1", "sha256": "8284731acc1558aba5b5d7f0514d64fcf3b2f67085a29afdfa8599cb26300ccc" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12289", "title": "TestPrism: Rethinking Test Evaluation Beyond a Single Reference", "published": "2026-10-08T16:44:39Z", "reviews": [ { "id": "2610.12289", "title": "TestPrism: Rethinking Test Evaluation Beyond a Single Reference", "published": "2026-10-08T16:44:39Z", "decision": "reviewed_deferred", "reason": "已读TestPrism构建/TestHelix方法：300task×10有效/无效implementation，joint success同时接受全valid拒绝全invalid，异构test-repair+peer crossvalidation+跨taskRSI。下一批需reference panel外部分数，单reference通过率不等价测试质量。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 TestPrism", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 TestHelix", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12289v1", "sha256": "1559d9239e24b10681196f256ab747d9696d0ca521f5a7e0dcd96034fb7be374", "local_html_sha256": "1559d9239e24b10681196f256ab747d9696d0ca521f5a7e0dcd96034fb7be374", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12289v1", "result": "retrieved", "sha256": "1559d9239e24b10681196f256ab747d9696d0ca521f5a7e0dcd96034fb7be374" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12281", "title": "Unlocking the Regulatory Genome by ARGUS: An Evidence-Constrained Agentic Framework for Interpreting Single Nucleotide Variants", "published": "2026-10-08T16:38:32Z", "reviews": [ { "id": "2610.12281", "title": "Unlocking the Regulatory Genome by ARGUS: An Evidence-Constrained Agentic Framework for Interpreting Single Nucleotide Variants", "published": "2026-10-08T16:38:32Z", "decision": "reviewed_deferred", "reason": "已读ARGUS架构/结果：458DNABERT binding模型、观察驱动planner与确定性evidence verifier分离，真实数据库query但仅单variant四TF case。科学agent候选保留，医疗解释不能推广单case，完整生物data/model依赖待审，暂不优先实施。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 Architecture", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12281v1", "sha256": "b5fb7b98df208c05dc8eda0b59db8d0dfbce2603f2cb64e3f6869bef10ddf76d", "local_html_sha256": "b5fb7b98df208c05dc8eda0b59db8d0dfbce2603f2cb64e3f6869bef10ddf76d", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12281v1", "result": "retrieved", "sha256": "b5fb7b98df208c05dc8eda0b59db8d0dfbce2603f2cb64e3f6869bef10ddf76d" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12274", "title": "HarnessSQL: Harness-Native Training for SQL Agents in Realistic Database Environments", "published": "2026-10-08T16:36:39Z", "reviews": [ { "id": "2610.12274", "title": "HarnessSQL: Harness-Native Training for SQL Agents in Realistic Database Environments", "published": "2026-10-08T16:36:39Z", "decision": "reviewed_deferred", "reason": "HarnessSQL在hermetic live DB/harness中教师verified全interaction SFT→execution RL，79db训练与100val，Spider2/BIRD/LiveSQL transfer；下一批必须hide execution oracle、readonly DB sandbox和4tool真实执行。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experimental Setup", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12274v1", "sha256": "fe5902a21b323d42333a4533b879121cafc8f2d75484f0374d1b3fcd7cb461ce", "local_html_sha256": "fe5902a21b323d42333a4533b879121cafc8f2d75484f0374d1b3fcd7cb461ce", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12274v1", "result": "retrieved", "sha256": "fe5902a21b323d42333a4533b879121cafc8f2d75484f0374d1b3fcd7cb461ce" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.12269", "title": "Cadence: Strategic Guidance for Coding Agents", "published": "2026-10-08T16:34:02Z", "reviews": [ { "id": "2610.12269", "title": "Cadence: Strategic Guidance for Coding Agents", "published": "2026-10-08T16:34:02Z", "decision": "reviewed_deferred", "reason": "已读Cadence方法和motivating study：advisory/replacement双层反馈驱动inspection频率，300SWE-Lite两harness。下一批需真实监控与agent执行/失败干预成本，固定interval或静态轨迹评分不实现adaptive scheduler。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3. Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2. Motivating Study", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12269v1", "sha256": "270df5098748625eb520630c1fe802d5ca53580bca521bacee5ad56c758069a1", "local_html_sha256": "270df5098748625eb520630c1fe802d5ca53580bca521bacee5ad56c758069a1", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12269v1", "result": "retrieved", "sha256": "270df5098748625eb520630c1fe802d5ca53580bca521bacee5ad56c758069a1" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12266", "title": "DVD: Dynamic Vector Decoding for Efficient MLLM-based Perception", "published": "2026-10-08T16:32:28Z", "reviews": [ { "id": "2610.12266", "title": "DVD: Dynamic Vector Decoding for Efficient MLLM-based Perception", "published": "2026-10-08T16:32:28Z", "decision": "reviewed_deferred", "reason": "DVD将2Dbox/mask/3Dbox编码1Dvectors，经learned1Dencoderdecoder+统一VQcodebook变少量perceptiontokens，真实MLLM输出这些tok再detokenize连续geometry。; RefCOCO系列和SUNRGBD/KITTI/Hypersim/nuScenes，2D/3Djointperception与token/latency；需所有codec/MLLM训练阶段。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12266v1#S3", "https://arxiv.org/html/2610.12266v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12266v1", "sha256": "794087b6cf7fadeadc59577053ac1b03872999ef1ca32b3896efda7e8ab00616" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12248", "title": "EgoVoice: Proactive Spoken Assistance from Egocentric Multimodal Streams", "published": "2026-10-08T16:24:12Z", "reviews": [ { "id": "2610.12248", "title": "EgoVoice: Proactive Spoken Assistance from Egocentric Multimodal Streams", "published": "2026-10-08T16:24:12Z", "decision": "reviewed_deferred", "reason": "EgoVoice用HoloAssist source-separated/resynth audio和listen-or-speak SFT/DPO训练主动语音助手，170 sessions onset/content/end评测；多模态/语音后训练下一批须实际流式音视频与timing policy。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12248v1", "sha256": "739fa088b70399874fa19e26e6d2ef62271c43adc60495b02c979cdeb5d7298a", "local_html_sha256": "739fa088b70399874fa19e26e6d2ef62271c43adc60495b02c979cdeb5d7298a", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12248v1", "result": "retrieved", "sha256": "739fa088b70399874fa19e26e6d2ef62271c43adc60495b02c979cdeb5d7298a" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.12242", "title": "TokenRouter: Efficient Serving System for Token-Level LLM Routing", "published": "2026-10-08T16:21:50Z", "reviews": [ { "id": "2610.12242", "title": "TokenRouter: Efficient Serving System for Token-Level LLM Routing", "published": "2026-10-08T16:21:50Z", "decision": "reviewed_deferred", "reason": "TokenRouter requestcentric route/send/receive coroutine，modelcentric subservers异步tri-loop/handoffresume，延后batchadmission阈值由DTMC吞吐模型据N/routingprob/step latency确定。; SGLangbackend五tokenrouting算法，Qwen0.6/8/32B、多pairs/workloads；全部并发、perrequest outputtokens/elapsed吞吐定义。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12242v1#S3", "https://arxiv.org/html/2610.12242v1#S4", "https://arxiv.org/html/2610.12242v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12242v1", "sha256": "ce243cfd9db0d74a91ecf2b6239dfd887158f004ecbe3ddee1545bda70157ae2" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12235", "title": "Language Models as AI Research World Models", "published": "2026-10-08T16:18:54Z", "reviews": [ { "id": "2610.12235", "title": "Language Models as AI Research World Models", "published": "2026-10-08T16:18:54Z", "decision": "reviewed_deferred", "reason": "Research World Models用真实实验记录预测unseen interventions并按预算选实验，九个pre/post/inference环境；适合auto-research controller下一批，必须保持outcome heldout、真实执行再验证，预测不能当已完成实验。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "2 Language Models as Research World Models", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "3 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12235v1", "sha256": "d1d254f14a7c1b85aeddbc1a5261848692d0d8ca16eea25086b1d061ddebad24", "local_html_sha256": "d1d254f14a7c1b85aeddbc1a5261848692d0d8ca16eea25086b1d061ddebad24", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12235v1", "result": "retrieved", "sha256": "d1d254f14a7c1b85aeddbc1a5261848692d0d8ca16eea25086b1d061ddebad24" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.12207", "title": "SciTBERT: A family of chronologically consistent language models for scientific and technological language processing", "published": "2026-10-08T15:57:45Z", "reviews": [ { "id": "2610.12207", "title": "SciTBERT: A family of chronologically consistent language models for scientific and technological language processing", "published": "2026-10-08T15:57:45Z", "decision": "reviewed_deferred", "reason": "SciTBERT年份2013-2025 cutoff pretrain/citation posttrain保持chronology，SciRepEval/PatRepEval embeddings任务；foundation encoder nextbatch需历史cutoff与citation pairs隔离，不能把现有晚期模型当无lookahead。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "Contrastive post-training with chronologically consistent citation data", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "Evaluation setup", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12207v1", "sha256": "ba6bd944317d96a918277a1e11fd0215ea6d2f1941cc6ceb7569146bb8b4bd90", "local_html_sha256": "ba6bd944317d96a918277a1e11fd0215ea6d2f1941cc6ceb7569146bb8b4bd90", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12207v1", "result": "retrieved", "sha256": "ba6bd944317d96a918277a1e11fd0215ea6d2f1941cc6ceb7569146bb8b4bd90" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.12201", "title": "SteerablePlex: Can We Steer Full-Duplex Models?", "published": "2026-10-08T15:56:28Z", "reviews": [ { "id": "2610.12201", "title": "SteerablePlex: Can We Steer Full-Duplex Models?", "published": "2026-10-08T15:56:28Z", "decision": "reviewed_deferred", "reason": "SteerablePlex full-duplex actor+asynchronous director，GDPO LoRA训练保turn-taking并跟多stage文本goal；200 SimIF scenarios/音频ASR-VAD评测，下一批需真实full-duplex streaming/director异步和rewarddecoupling。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiment", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12201v1", "sha256": "c30f3cb3c91006da12cb43ababc3531a23fc30dd70faf972921ff345abf21edb", "local_html_sha256": "c30f3cb3c91006da12cb43ababc3531a23fc30dd70faf972921ff345abf21edb", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12201v1", "result": "retrieved", "sha256": "c30f3cb3c91006da12cb43ababc3531a23fc30dd70faf972921ff345abf21edb" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.12183", "title": "A Closer Look at Agentic BBO: Benchmarking LLM Agents for Black-Box Optimization", "published": "2026-10-08T15:48:07Z", "reviews": [ { "id": "2610.12183", "title": "A Closer Look at Agentic BBO: Benchmarking LLM Agents for Black-Box Optimization", "published": "2026-10-08T15:48:07Z", "decision": "reviewed_deferred", "reason": "已读AgenticBBO benchmark设计/实验：五域有限预算、统一family score、numerical optimizer/tool/prior/LLM角色消融，公开code报告。下一批需真实expensive objective与验证选择，不能把test optimum/gold函数信息给优化agent。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Benchmark Design", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12183v1", "sha256": "68c96b8dd298a285b45992787aee3a12740b08504e8493756d0f9095e22527b4", "local_html_sha256": "68c96b8dd298a285b45992787aee3a12740b08504e8493756d0f9095e22527b4", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12183v1", "result": "retrieved", "sha256": "68c96b8dd298a285b45992787aee3a12740b08504e8493756d0f9095e22527b4" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12147", "title": "Healthy Counterfactual Generation via Diffusion Inpainting for Mammography Classification", "published": "2026-10-08T15:31:56Z", "reviews": [ { "id": "2610.12147", "title": "Healthy Counterfactual Generation via Diffusion Inpainting for Mammography Classification", "published": "2026-10-08T15:31:56Z", "decision": "rejected_or_out_of_scope", "reason": "在健康 BI-RADS 1 乳腺图上训练 DDPM，再用 RePaint 在已标注病灶框内重绘正常组织；保留框外结构，并把生成图作为负类训练增广，而非训练通用基础模型。; VinDr-Mammo 5000 检查/20000 图像，官方16000训练、4000测试；512缩放，训练集增加1410健康反事实。ConvNext、ViT、MammoCLIP、FPN-MIL敏感度多数改善，但特异度下降且部分ROC-AUC下降；另报告固定80%特异度下召回及放射科医师真实性判断。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12147v1#S2", "https://arxiv.org/html/2610.12147v1#S3", "https://arxiv.org/html/2610.12147v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12147v1", "sha256": "2f5918f138b62b8d45ccb3177331859db70eeaf003c4d05f4b0f5f0ad74e4f1c" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.12144", "title": "Language-Specific Effects of Tokenizer Choice in Multilingual Language Models", "published": "2026-10-08T15:29:37Z", "reviews": [ { "id": "2610.12144", "title": "Language-Specific Effects of Tokenizer Choice in Multilingual Language Models", "published": "2026-10-08T15:29:37Z", "decision": "reviewed_deferred", "reason": "以同一nanochat d24 1.27B架构、训练语料/配方/词表约128K比较54种tokenizer、123个LM；按语言分析BPB对tokenizer配置的敏感性，并做单因素配置和tokenizer训练语言配比重分配。; 31训练语言，28有词边界；每语言BPB变化与LM训练份额负相关。英语独占tokenizer训练损害几乎全部语言，equal/inverse配比不必改善低资源语言，尤其LM训练重复数据时。XNLI/MultiBLiMP有区分，XCOPA/Belebele在该规模近随机。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12144v1#S2", "https://arxiv.org/html/2610.12144v1#S3", "https://arxiv.org/html/2610.12144v1#S4", "https://arxiv.org/html/2610.12144v1#S5", "https://arxiv.org/html/2610.12144v1#S8" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12144v1", "sha256": "3b10a00bd6ef20bdc3259a6151ed6724533bc3c80cdaec5944f588b3744a0ea3" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12135", "title": "Q-Shaped Options for Hierarchical Reinforcement Learning", "published": "2026-10-08T15:23:39Z", "reviews": [ { "id": "2610.12135", "title": "Q-Shaped Options for Hierarchical Reinforcement Learning", "published": "2026-10-08T15:23:39Z", "decision": "rejected_or_out_of_scope", "reason": "已读完整摘要，官方正文不可用：Q-Shaped Options用高低层Q共同shape option encoder，offline locomotion/manipulation传统HRL非LLM，本次agent批范围外；不宣称已核验全文实验。", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "read_scope": "abstract_only_primary_fulltext_unavailable", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "url": "https://arxiv.org/abs/2610.12135", "sha256": null, "local_html_sha256": null, "local_html_sha256_matches": false, "attempts": [ { "url": "https://arxiv.org/html/2610.12135v1", "result": "failed", "error": "404 Client Error: Not Found for url: https://arxiv.org/html/2610.12135v1" }, { "url": "https://arxiv.org/html/2610.12135v1", "result": "failed", "error": "404 Client Error: Not Found for url: https://arxiv.org/html/2610.12135v1" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.12133", "title": "Rehearse Everything, Remember Nothing: Attic-KV Rehearses What Will Be Read", "published": "2026-10-08T15:22:50Z", "reviews": [ { "id": "2610.12133", "title": "Rehearse Everything, Remember Nothing: Attic-KV Rehearses What Will Be Read", "published": "2026-10-08T15:22:50Z", "decision": "reviewed_deferred", "reason": "Attic-KV在未知未来query时，用模型自身生成的、引用上下文内容的QA及内容自适应anchor token来rehearsal后压缩KV；作为KVzip/KVgrad/RestoreKV+的rehearsal替换，不把未来测试query送入缓存选择。; Qwen3-4B/8B/14B及Llama系列，RULER和LongBench；比较不同keep ratio、固定比例rehearsal与compress-once answer-many，并报告压缩成本。650道RULER开发问题与全部测试隔离。改善因类别/host方法而异，少数multi-doc QA项持平或稍差。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12133v1#S3", "https://arxiv.org/html/2610.12133v1#S4", "https://arxiv.org/html/2610.12133v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12133v1", "sha256": "ef293f6b2b9913c4557855dfd101748786badeb41dedfda98df87db0600ba287" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12126", "title": "SuperNav: An Agentic Navigation System for Any Task in Any Scene", "published": "2026-10-08T15:19:40Z", "reviews": [ { "id": "2610.12126", "title": "SuperNav: An Agentic Navigation System for Any Task in Any Scene", "published": "2026-10-08T15:19:40Z", "decision": "reviewed_deferred", "reason": "SuperNav以冻结MLLM为导航规划器，通过Skills说明、像素点视觉接口和真实导航工具构成harness；执行器提供几何/学习式底层控制，任务进度与上下文循环管理高层计划。; Habitat-GS、AI2THOR、HM3D比较GPT-5.6 Terra加不同执行器；另有Unitree Go2四视角RGB+LiDAR实机对象寻找演示。接口与历史输入和原生baseline不完全一致，实机演示不是生产A/B。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12126v1#S3", "https://arxiv.org/html/2610.12126v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12126v1", "sha256": "8f3f220afabf303cf5d0e5b44c2446b4524f38457c2cd62e2b41e21b96eaf037" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12124", "title": "Use and Disuse: Intent-Structured Experience Consolidation for Memory and Learning in LLM Agents", "published": "2026-10-08T15:16:56Z", "reviews": [ { "id": "2610.12124", "title": "Use and Disuse: Intent-Structured Experience Consolidation for Memory and Learning in LLM Agents", "published": "2026-10-08T15:16:56Z", "decision": "reviewed_deferred", "reason": "已读Hippocam概览/实验：nested intent、Intent/Prefix consolidation、保留raw interaction层级展开、recall后与新experience再consolidate，无参数更新。下一批/已有相关实现对照须保留多LLM组件，不用标签/固定summary充当机制。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 Hippocam", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12124v1", "sha256": "ab528d62b906f10ae89de18bab06d70492a572289cc592578860c82d3ec4c42a", "local_html_sha256": "ab528d62b906f10ae89de18bab06d70492a572289cc592578860c82d3ec4c42a", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12124v1", "result": "retrieved", "sha256": "ab528d62b906f10ae89de18bab06d70492a572289cc592578860c82d3ec4c42a" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12086", "title": "EvoAlloc: A Self-Evolving Resource Allocation Agent for Efficient Program Evolution", "published": "2026-10-08T14:57:35Z", "reviews": [ { "id": "2610.12086", "title": "EvoAlloc: A Self-Evolving Resource Allocation Agent for Efficient Program Evolution", "published": "2026-10-08T14:57:35Z", "decision": "reviewed_deferred", "reason": "已读EvoAlloc方法/实验：explicit allocation strategy按search experience自修订，定期经验consolidation和counterfactual抽评被拒候选纠正censored feedback；ShinkaEvolve五seed matched-target评测。下一批真实候选与成本，不用已实现fitness当forecast。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12086v1", "sha256": "ca882dbef3a37eaf5c70266c07e0e47957b3222dc9d90668b33fb163ae4c7172", "local_html_sha256": "ca882dbef3a37eaf5c70266c07e0e47957b3222dc9d90668b33fb163ae4c7172", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12086v1", "result": "retrieved", "sha256": "ca882dbef3a37eaf5c70266c07e0e47957b3222dc9d90668b33fb163ae4c7172" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12061", "title": "When Should Agents Think? Adaptive Reasoning via Cross-Turn Estimation", "published": "2026-10-08T14:43:09Z", "reviews": [ { "id": "2610.12061", "title": "When Should Agents Think? Adaptive Reasoning via Cross-Turn Estimation", "published": "2026-10-08T14:43:09Z", "decision": "accepted", "reason": "RACE/LoGiC通过移除reasoning后的reference-action likelihood检测跨turn支持，结合SFT/RL；已有implemented条目，本次保留并核对原文，不重复宣称新实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12061v1", "sha256": "077e07c12d1ac7fc5d819c34ca359357e34ba2e74887cc097ac40206119b7ba6", "local_html_sha256": "077e07c12d1ac7fc5d819c34ca359357e34ba2e74887cc097ac40206119b7ba6", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12061v1", "result": "retrieved", "sha256": "077e07c12d1ac7fc5d819c34ca359357e34ba2e74887cc097ac40206119b7ba6" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "post-training" ], "status": "accepted" }, { "id": "2610.12060", "title": "Look Back, Think Ahead: Visual Memory on Demand for Efficient Multimodal Reasoning", "published": "2026-10-08T14:42:57Z", "reviews": [ { "id": "2610.12060", "title": "Look Back, Think Ahead: Visual Memory on Demand for Efficient Multimodal Reasoning", "published": "2026-10-08T14:42:57Z", "decision": "reviewed_deferred", "reason": "ViMoD在冻结Qwen3-VL-4B中保留全部原始Fine视觉/text KV，DART生成内容自适应Coarse groups及容量；TRACE随解码历史执行SELECT/HOLD/REPLACE Fine group。教师预告将需区域仅用于训练，学生轨迹用RLOO正确性与成功条件下occupancy目标。; 20%目标活动视觉预算，8个bench，新增训练参数约0.0546%；报告活动访问/总耗时下降。峰值内存42.67GiB高于Full的38.40GiB且跨预算基本不变，因为完整Fine/Coarse bank一直保留。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12060v1#S3", "https://arxiv.org/html/2610.12060v1#S4", "https://arxiv.org/html/2610.12060v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12060v1", "sha256": "fbbd09f31ae69938478851cee41da685f68208ddf5712f09dac7c8951aed4d6f" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12048", "title": "Do Not Train Away Uncertainty: Early Uncertainty Anchored Calibration", "published": "2026-10-08T14:34:26Z", "reviews": [ { "id": "2610.12048", "title": "Do Not Train Away Uncertainty: Early Uncertainty Anchored Calibration", "published": "2026-10-08T14:34:26Z", "decision": "reviewed_deferred", "reason": "EUA-Cal冻结早期checkpoint作为不确定性锚点；EPR对早期预测和当前预测做KL，PSR用早期真实类置信度加权归一化特征构造固定class prototype，再将余弦相似度经temperature变为soft target、对当前预测做KL，与CE联合训练。; CIFAR10/100、TinyImageNet多种CNN/ViT；SciQ、ARC-E/C、CSQA上Llama3/3.1-8B与Qwen3-8B-Base LoRA。比较ACC/ECE/AECE与OOD AUROC/FPR95及校准baselines，报告tau变化分析。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.12048v1#S3", "https://arxiv.org/html/2610.12048v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12048v1", "sha256": "bd0b8b91ffa6e09d3b83e1be328828e40f6f6f03dbf70de4105505f0a6ac94a7" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12022", "title": "Examining Social Attribution in LLM Reasoning: A Theory-Guided Probing Methodology", "published": "2026-10-08T14:20:06Z", "reviews": [ { "id": "2610.12022", "title": "Examining Social Attribution in LLM Reasoning: A Theory-Guided Probing Methodology", "published": "2026-10-08T14:20:06Z", "decision": "rejected_or_out_of_scope", "reason": "已读social attribution方法/实验：7639责任/归责问题、32LLM及latent probe研究人类归因理论一致性，非工具agent执行或workflow算法，本次agent实现范围外的foundation行为研究。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12022v1", "sha256": "e73cf684e41d39b3ccff27cdfb5bca0907368e94af5841c96dfa0bce798cc520", "local_html_sha256": "e73cf684e41d39b3ccff27cdfb5bca0907368e94af5841c96dfa0bce798cc520", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12022v1", "result": "retrieved", "sha256": "e73cf684e41d39b3ccff27cdfb5bca0907368e94af5841c96dfa0bce798cc520" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.12002", "title": "Agentic-TTT: Training test-time policy for test-time training", "published": "2026-10-08T14:06:26Z", "reviews": [ { "id": "2610.12002", "title": "Agentic-TTT: Training test-time policy for test-time training", "published": "2026-10-08T14:06:26Z", "decision": "reviewed_deferred", "reason": "已读Agentic-TTT问题/实验：learnedpolicy决定是否及哪种TTT、skill复用，真实TTT tools和library状态由utility训练，187train/176heldout stream。下一批必须真实权重更新工具链并防test gold被优化policy读取，utility/compute权衡不可heuristic替代。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 Agentic Test-Time Training", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.12002v1", "sha256": "41bcbd25c5869eadaeadaca70ab1e516b5432c3ecb18918cf343c1764ce46bbe", "local_html_sha256": "41bcbd25c5869eadaeadaca70ab1e516b5432c3ecb18918cf343c1764ce46bbe", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12002v1", "result": "retrieved", "sha256": "41bcbd25c5869eadaeadaca70ab1e516b5432c3ecb18918cf343c1764ce46bbe" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11993", "title": "DataVista: Diagnosing Multimodal LLMs on Data Video Understanding", "published": "2026-10-08T14:02:35Z", "reviews": [ { "id": "2610.11993", "title": "DataVista: Diagnosing Multimodal LLMs on Data Video Understanding", "published": "2026-10-08T14:02:35Z", "decision": "reviewed_deferred", "reason": "DataVista从真实数据叙事视频构造三级data perception/temporal reasoning/narrative understanding及10类问题；人机协作问题生成、文本过滤后专家回视频核验/修改歧义，并保留多选精确集合评分。; 961视频、6775问题、19MLLM；主评统一50均匀帧无音频，有/无subtitle同一帧，overall合并两条件。单选exact match、多选必须选项集合完全正确。最佳Gemini3.1Pro70%，另诊断帧数/subtitle及chart reading、causal/narrative难项。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11993v1#S3", "https://arxiv.org/html/2610.11993v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11993v1", "sha256": "c3a977f1c133915c94f4179d5a9a99e8cbec97f58d58470d69894a1beed29e88" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11986", "title": "FearCaut-Qwen: Affective Steering in a Vision-Language Model Shifts the Decision Criterion for Hazard Assessment", "published": "2026-10-08T13:57:46Z", "reviews": [ { "id": "2610.11986", "title": "FearCaut-Qwen: Affective Steering in a Vision-Language Model Shifts the Decision Criterion for Hazard Assessment", "published": "2026-10-08T13:57:46Z", "decision": "reviewed_deferred", "reason": "在独立FindingEmo discovery上末prompt-token residual逐层linear probe、validation选L22-24；计算fear-vs-neutral classmean residual direction，validation选非退化alpha、内容hash冻结后在Qwen2.5VL-7B两阶段SeisMLLM perception/retrieval/posting中注入。另SwiGLU sparse neuron knockout用matched controls验证源情绪因果必要性，和distributed方向分别检验迁移。; 官方65建筑test不变，train-fit1039/validation202；emotion844discovery/280validation/280test ID互斥。fear alpha+.5 Red recall .270→.757但d'1.521→1.028、criterion+1.354→-.161；perception-only重现94.4%效应，sparse neurons target迁移无效。配对case bootstrap10000次及多重比较校正。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "official PDF selected page prefixes (explicit ranges below), not complete48pages", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/pdf/2610.11986v1#page=7", "https://arxiv.org/pdf/2610.11986v1#page=16", "https://arxiv.org/pdf/2610.11986v1#page=20", "https://arxiv.org/pdf/2610.11986v1#page=23", "https://arxiv.org/pdf/2610.11986v1#page=29", "https://arxiv.org/pdf/2610.11986v1#page=32", "https://arxiv.org/pdf/2610.11986v1#page=33" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/pdf/2610.11986v1", "sha256": "8df0a7c6aeafa1b3033802c88c1ce6ebdbcc5cfc54e1705e8737ddee761b91e1" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11966", "title": "MindFlow: Mind Supernet Powered Thinking Flows for Research Idea Innovation", "published": "2026-10-08T13:45:51Z", "reviews": [ { "id": "2610.11966", "title": "MindFlow: Mind Supernet Powered Thinking Flows for Research Idea Innovation", "published": "2026-10-08T13:45:51Z", "decision": "reviewed_deferred", "reason": "已读MindFlow问题与IdeaBench构建：probabilistic mind supernet/controller以tournament相对rank优化thinkingflow；benchmark由target paper/source构成。下一批需真实controller学习与独立novelty/feasibility评测，防target方案进入生成输入，LLM rank非科学验证。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "5 Benchmark Construction", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Problem Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11966v1", "sha256": "7947aa008ba129f233f8f4a43ba4d161abd3babcc1db36755a237dc9d4169f8c", "local_html_sha256": "7947aa008ba129f233f8f4a43ba4d161abd3babcc1db36755a237dc9d4169f8c", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11966v1", "result": "retrieved", "sha256": "7947aa008ba129f233f8f4a43ba4d161abd3babcc1db36755a237dc9d4169f8c" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11963", "title": "Can LLMs Fix It Without Code? Toward Automated Verification of No-Code Bug Fixes", "published": "2026-10-08T13:44:53Z", "reviews": [ { "id": "2610.11963", "title": "Can LLMs Fix It Without Code? Toward Automated Verification of No-Code Bug Fixes", "published": "2026-10-08T13:44:53Z", "decision": "reviewed_deferred", "reason": "已读no-code verification方法/结果：rollback浏览器runtime、issue-specific checker与setup sanity gates，3 executor替换会显著改变resolution；仅17.6%issues可复现。下一批评测infra要独立控制executor、人工agreement，不把applicability当修复成功。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3. Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4. Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11963v1", "sha256": "9fad44164a6f00446d2c91a31e3e77a4aab8145ac3747e9d1ce0165c1aebda95", "local_html_sha256": "9fad44164a6f00446d2c91a31e3e77a4aab8145ac3747e9d1ce0165c1aebda95", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11963v1", "result": "retrieved", "sha256": "9fad44164a6f00446d2c91a31e3e77a4aab8145ac3747e9d1ce0165c1aebda95" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11959", "title": "MiMo-V2.6: Scaling Reinforcement Learning Towards Self-Improvement", "published": "2026-10-08T13:41:53Z", "reviews": [ { "id": "2610.11959", "title": "MiMo-V2.6: Scaling Reinforcement Learning Towards Self-Improvement", "published": "2026-10-08T13:41:53Z", "decision": "reviewed_deferred", "reason": "MiMo-V2.6真实大规模agentic RL：1568 prompts×16 rollouts，asynchronous GRPO/staleness4、frozen MoE router及groupwise graders；公开训练日志/环境链路，资源规模远超冻结批次，下一批需要分解可执行机制。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "4 Scaling Reinforcement Learning", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments: You Only RL Once", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11959v1", "sha256": "b07bc0b1e19ddc8bd6d8753fadbf9b2c5c3eb7899f7925650f1ff1e81df963aa", "local_html_sha256": "b07bc0b1e19ddc8bd6d8753fadbf9b2c5c3eb7899f7925650f1ff1e81df963aa", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11959v1", "result": "retrieved", "sha256": "b07bc0b1e19ddc8bd6d8753fadbf9b2c5c3eb7899f7925650f1ff1e81df963aa" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11954", "title": "Look Where You Can: Active View Selection for CAD Reconstruction under Occlusion", "published": "2026-10-08T13:40:38Z", "reviews": [ { "id": "2610.11954", "title": "Look Where You Can: Active View Selection for CAD Reconstruction under Occlusion", "published": "2026-10-08T13:40:38Z", "decision": "reviewed_deferred", "reason": "SightCAD以冻结DINOv2图像特征+pose编码+cross-view transformer从完整可行候选view集选择K；warm start蒸馏SA-max与VLM多视角CadQuery SFT；随后selector用生成/执行后voxelIoU的GRPO、generator在组内最高reward views做GT-code SFT；最后固定selector独立generator GRPO，避免两policy信用混淆。; CAD-Recode/Text2CAD约4k heldout，冻结相同场景placement/occlusion；DeepCAD/Fusion360及T-LESS、MP6D synthetic/real。报告invalid code、valid-onlyChamfer/mIoU及invalid记0的effective mIoU；viewpolicy按固定metric尺度，cross-method各自unitcube归一化。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11954v1#S3", "https://arxiv.org/html/2610.11954v1#S4", "https://arxiv.org/html/2610.11954v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11954v1", "sha256": "688d78a7fbbb48215e1667dc9cf17ee0c58eae2ccb3252ed6dc1631a2e7a7472" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11948", "title": "When History Helps and Hurts: Selective History Use across Multimodal Turns", "published": "2026-10-08T13:36:12Z", "reviews": [ { "id": "2610.11948", "title": "When History Helps and Hurts: Selective History Use across Multimodal Turns", "published": "2026-10-08T13:36:12Z", "decision": "reviewed_deferred", "reason": "ReTurn将任务携带历史Reconfirm/Reground与证据携带历史Retrieve/Rebind做成保持target question/media/answer的配对局部对照，实际模型生成历史；配套restatement、基于显式reference的attention reweighting及Short OpenQA SFT/answer-preserving JS penalty。; 7000base任务、视觉/音频、13模型matched singleturn/multiturn；median OpenQA93.7%到72.3%。Qwen3Omni单seed DevTrain509pairs/1018views、204updates；SFT+3.42pp OpenQA但LongMCQ Reconfirm-1.3pp，Inv仅+.02/.20pp且interval跨0。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11948v1#S3", "https://arxiv.org/html/2610.11948v1#S4", "https://arxiv.org/html/2610.11948v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11948v1", "sha256": "fcd027e5be73654a76c0bf7b380e30d27798ca451576bf36acf4995e11ddfe74" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11938", "title": "Does Target Alignment Mean Target Recovery? An Evidence-Ladder Study of Adversarial Claims on Contrastive Encoders", "published": "2026-10-08T13:30:52Z", "reviews": [ { "id": "2610.11938", "title": "Does Target Alignment Mean Target Recovery? An Evidence-Ladder Study of Adversarial Claims on Contrastive Encoders", "published": "2026-10-08T13:30:52Z", "decision": "reviewed_deferred", "reason": "将contrastive encoder目标攻击证据分成L0intent/L1victimcosinealignment/L2独立模型recovery/L3downstream/L4human。用训练/架构不同的supervised ResNet50、补充ViTjudge，真实target图blend正控、shuffledtarget负控；目标logitmargin变化减16个shuffle变化为ITE，预冻结50%blend reference后confirmatory匹配攻击。; 六encoder、三扰动预算，robust FARE/TeCoA/PMG/TRADES独立证据高于vanilla CLIP/SigLIP，但无cell达到blend-derived reference，前三在replication band故实际recovery仍未定。robust内部alignment/ITErho .24-.51，vanilla CLIP近0，跨encoder无单调关系。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11938v1#S3", "https://arxiv.org/html/2610.11938v1#S4", "https://arxiv.org/html/2610.11938v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11938v1", "sha256": "5d77aa12091435d49aa409f3703ba6838f586d3328f26482dd45eb4d50376a05" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11924", "title": "Revisiting Identity and Spectra Dispersion in Media-Bridged Time Series Forecasting: Linking Multivariate Signals and Narrative Flows", "published": "2026-10-08T13:20:49Z", "reviews": [ { "id": "2610.11924", "title": "Revisiting Identity and Spectra Dispersion in Media-Bridged Time Series Forecasting: Linking Multivariate Signals and Narrative Flows", "published": "2026-10-08T13:20:49Z", "decision": "reviewed_deferred", "reason": "MIDAPN先以frozen PLM将预对齐叙事变为数值pseudo-variates，再DFF频域融合、MIDAG static essence/dynamic behavior/latentcommonality及CIM图聚合。SPConv对FFT实/虚系数做局部卷积、层级avgpool压缩与通道扩张，按temporal reconstructioncapacity约束对整数expansion进行mixed-order搜索。; 13multivariate和12预对齐numerical-text数据；16native TSF与14TSFM/fusedPLM，TaTS统一prealignment，官方描述ETT/PEMS6:2:2、其他7:1:2，MAE/MSE；另fixedlongcontext比较不同预训练TSFM。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11924v1#S3", "https://arxiv.org/html/2610.11924v1#S4", "https://arxiv.org/html/2610.11924v1#S5", "https://arxiv.org/html/2610.11924v1#S14" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11924v1", "sha256": "677550be2e6a925e92fe1d5fa8f4fe69d0db7028665ab6d2e19fc01e0811fe44" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11920", "title": "Event-Centric Memory with Query-Aware Graph Augmentation for Long-Term Conversational Agents", "published": "2026-10-08T13:17:59Z", "reviews": [ { "id": "2610.11920", "title": "Event-Centric Memory with Query-Aware Graph Augmentation for Long-Term Conversational Agents", "published": "2026-10-08T13:17:59Z", "decision": "reviewed_deferred", "reason": "事件原子记忆和动态trace保存来源/当前与历史状态；hybrid retrieval+query rerank激活局部graph，关系感知GNN mean pooling+projector成为真实graph token供LLM。; LoCoMo、LongMemEval-S、MemoryAgentBench conflict、BEAM到10M histories；judge重复10次，并有retrieval和预算敏感性。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11920v1#S3", "https://arxiv.org/html/2610.11920v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11920v1", "sha256": "ae6134dcf1fac694d8d7fb80a7ab98bfb78bd1afe01f78d4467abc3724f6ea14" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11907", "title": "Beyond Visual Enhancement: Adaptive Multi-Context Steering to Mitigate LVLM Hallucinations", "published": "2026-10-08T13:08:44Z", "reviews": [ { "id": "2610.11907", "title": "Beyond Visual Enhancement: Adaptive Multi-Context Steering to Mitigate LVLM Hallucinations", "published": "2026-10-08T13:08:44Z", "decision": "reviewed_deferred", "reason": "AIMS在native per-head query/key空间对视觉与prefill文本keycache均值prototype、生成context用最新生成token key作为summary；((cos(q,p)+1)/2)^(1/tau_source)给head/timestep/source自适应权重，目标s为加权prototype和，qtilde=(1-alpha)q+alpha*s，然后原attention/KV不变、无辅助forward。; LLaVA1.5-7B/Qwen2.5VL-3B等，CHAIR500COCO初步分析与greedy/beam/nucleus主比较，多source ablation；报告CHAIRs/i、objectrecall/F1、captionlength及一般能力。CHAIR减少有部分recall/length下降，nucleus部分项收益较小。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11907v1#S3", "https://arxiv.org/html/2610.11907v1#S4", "https://arxiv.org/html/2610.11907v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11907v1", "sha256": "63e6c30d708c464964876289717f42ffebaf63ef77802b3012fcab08954a96d4" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11899", "title": "Forms of LLM-Integrated Applications from LLM-Chats to Autonomous AI Agent System", "published": "2026-10-08T13:05:33Z", "reviews": [ { "id": "2610.11899", "title": "Forms of LLM-Integrated Applications from LLM-Chats to Autonomous AI Agent System", "published": "2026-10-08T13:05:33Z", "decision": "rejected_or_out_of_scope", "reason": "描述22研究/厂商系统的调用控制、agent数、tool和人工介入的四维架构词汇；不是提出可执行的新能力算法。; illustrative corpus与单作者架构编码，无模型受控能力训练/benchmark。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11899v1#S3", "https://arxiv.org/html/2610.11899v1#S4", "https://arxiv.org/html/2610.11899v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11899v1", "sha256": "8b10da673f7004c723b180ef63227d6f4c8801980c11cfc648c9eeab874d6f42" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.11896", "title": "Automated Assembly Instruction Generation from CAD Models Using Grounded Large Language Models: A Human-in-the-Loop Framework", "published": "2026-10-08T13:04:03Z", "reviews": [ { "id": "2610.11896", "title": "Automated Assembly Instruction Generation from CAD Models Using Grounded Large Language Models: A Human-in-the-Loop Framework", "published": "2026-10-08T13:04:03Z", "decision": "rejected_or_out_of_scope", "reason": "STEP assembly转typedProductGraph，Kahn确定拓扑次序，LLM仅实现selectedgraphslice为自然语言，规则/LLM review标flag；human清完flag才允许PDF。现prototype几何仅bbox/containment，matingconstraints来自提供的assemblystructure。; 仅内置六part/八constraint referenceassembly单次proof-of-concept，Pillow示意图fallback；展示数量/材料/2.5Nm扭矩和次序保持。没有多assembly、geometrycollision验证、泛化或foundation model方法比较。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11896v1#S3", "https://arxiv.org/html/2610.11896v1#S4", "https://arxiv.org/html/2610.11896v1#S6", "https://arxiv.org/html/2610.11896v1#S8" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11896v1", "sha256": "d986b6ab62df897ad20969b2d3e4d087f09811f7742e89c56ed26bf992b51a29" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.11862", "title": "LEVER: Adaptive Cost-Aware Proof Search Over AND/OR Graphs", "published": "2026-10-08T12:40:14Z", "reviews": [ { "id": "2610.11862", "title": "LEVER: Adaptive Cost-Aware Proof Search Over AND/OR Graphs", "published": "2026-10-08T12:40:14Z", "decision": "reviewed_deferred", "reason": "已读LEVER AND/OR graph与实验：partial proof realized objective加open subgoal预测，programmable cost/length/impurity，Lean kernel保correct；matched model/version/price。下一批须真实Lean搜索、预测前评分与cost含失败，不能gold proof排序代替。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "5 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Lever", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11862v1", "sha256": "8b9f4ed62b4c4d831a170b2b148f6fae1c578c6b646513b4254a1d53ce4ef53c", "local_html_sha256": "8b9f4ed62b4c4d831a170b2b148f6fae1c578c6b646513b4254a1d53ce4ef53c", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11862v1", "result": "retrieved", "sha256": "8b9f4ed62b4c4d831a170b2b148f6fae1c578c6b646513b4254a1d53ce4ef53c" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11858", "title": "Trajectory-Guided Fault Localization for Agent Skill Evolution", "published": "2026-10-08T12:38:51Z", "reviews": [ { "id": "2610.11858", "title": "Trajectory-Guided Fault Localization for Agent Skill Evolution", "published": "2026-10-08T12:38:51Z", "decision": "reviewed_deferred", "reason": "已读SkillMorph方法/实验：多次真实轨迹抽象，对成败action coverage并结合跨loop变化做fault localization，定位skill edit再试；6PR接受是工业case不是生产A/B。下一批必须实际执行修订及验证集选择，不接受固定技能fixtures。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Approach", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiment", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11858v1", "sha256": "e72f300350309836219e42ef1fba4e42578df377a4e7123441fe339546d488c6", "local_html_sha256": "e72f300350309836219e42ef1fba4e42578df377a4e7123441fe339546d488c6", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11858v1", "result": "retrieved", "sha256": "e72f300350309836219e42ef1fba4e42578df377a4e7123441fe339546d488c6" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11854", "title": "GRPODropout: Less is More for Online Reinforcement Learning Rollouts", "published": "2026-10-08T12:38:17Z", "reviews": [ { "id": "2610.11854", "title": "GRPODropout: Less is More for Online Reinforcement Learning Rollouts", "published": "2026-10-08T12:38:17Z", "decision": "accepted", "reason": "GRPODropout已有实现：删高概率positive-advantage rollout并recenter retained credit，matched budget math/generalQA三scale比较；保留existing范围，不因survey再称本批新增。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11854v1", "sha256": "0c0f923d5467463326ae73186b4159a965ceb9358e7b6cf8ef011ee1080d8d43", "local_html_sha256": "0c0f923d5467463326ae73186b4159a965ceb9358e7b6cf8ef011ee1080d8d43", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11854v1", "result": "retrieved", "sha256": "0c0f923d5467463326ae73186b4159a965ceb9358e7b6cf8ef011ee1080d8d43" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "accepted" }, { "id": "2610.11826", "title": "From Suppression to Repair: Mitigating Object Hallucination in Large Vision-Language Models via Localized Distribution Alignment", "published": "2026-10-08T12:25:50Z", "reviews": [ { "id": "2610.11826", "title": "From Suppression to Repair: Mitigating Object Hallucination in Large Vision-Language Models via Localized Distribution Alignment", "published": "2026-10-08T12:25:50Z", "decision": "reviewed_deferred", "reason": "ResOT用LURE pairedfaithful/hallucinated完整caption末位hidden作离线calibration；两类SVDtop-r，hallucinatedbasis投影除faithfuldominantsubspace后SVD得residualU。对U内两类均值/收缩cov估计GaussianOT闭式affine；推理logdensityratio sigmoid gate控制alpha和normclip，h+=alpha*g*U(T(U^Th)-U^Th)，正交分量不变。; 5000LURE训练pairedcaptions，7B LLaVA1.5/mPLUGOwl2/InstructBLIP greedy；POPE、CHAIR/BLEU1、MME、MMHal/AMBER与一般multimodal任务，GPT4o评分COCO描述；不同校准样本量、image-text mismatch/labelswap腐化鲁棒性。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11826v1#S3", "https://arxiv.org/html/2610.11826v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11826v1", "sha256": "458bac4700cebeda1f12f9b778d8a352cb6dce39ccb0a82a09106628812db608" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11818", "title": "From Pixels to Structure: Lightweight Vision-Language Models for Document OCR and Structured JSON Extraction", "published": "2026-10-08T12:21:52Z", "reviews": [ { "id": "2610.11818", "title": "From Pixels to Structure: Lightweight Vision-Language Models for Document OCR and Structured JSON Extraction", "published": "2026-10-08T12:21:52Z", "decision": "reviewed_deferred", "reason": "比较八个200M-7B open VLM真实documentimage→OCR/schemaJSON。按各模型支持接口固定OCRprompt或JSON skeleton，自由prompt模型fewshot仅train示例；<1B全参、较大模型QLoRA。另独立Optuna validationCER调参、经典图像预处理、同epochbudget两阶段learningrate训练与jointcollections。; 三德国高校collection共1062records，743train/159val/160test；Erlangen308图公开archive、Gemini初稿人工核转写Zenodo。CER/ANLS*无效JSON回落rawtext，mAPF1无效JSON记0；最佳checkpoint最低autoregressive valCER。jointtraining出现数据集间性能转移。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11818v1#S3", "https://arxiv.org/html/2610.11818v1#S4", "https://arxiv.org/html/2610.11818v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11818v1", "sha256": "34e7a06e3091ed19d3713d9df6b2af428f5424e61c4937d47b6f1fd2065d6b08" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11810", "title": "Seek-and-View Reasoning for Multi-View Spatial Understanding", "published": "2026-10-08T12:18:01Z", "reviews": [ { "id": "2610.11810", "title": "Seek-and-View Reasoning for Multi-View Spatial Understanding", "published": "2026-10-08T12:18:01Z", "decision": "reviewed_deferred", "reason": "Vantage trainingfree VLM根据问题/原视图选reference、解释directionframe并从17相机action×3magnitude规划；G3T gravity-aligned3D reconstruction/pose将语义action映射metricpose并reproject合成question-relevantview。最终VQA同时保留原图、新图、analysis和guide，不用模型假想图替代真实3D几何。; 六VLM、5空间bench子集：MindCubeTiny、MMSI position/attribute、BLINK multiview，以及OmniSpatial perspective/SPIN rotationtranslation；H200×2 G3T与openVLM，accuracy与componentablation。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11810v1#S3", "https://arxiv.org/html/2610.11810v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11810v1", "sha256": "8e3047e24aa359a0014b855286b8ab6bed02789c0934a59f9aa9f7fc37bf133c" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11794", "title": "Memento 3: Model-Based Recursive Self-Improvement through Reflective Rulebooks", "published": "2026-10-08T12:06:54Z", "reviews": [ { "id": "2610.11794", "title": "Memento 3: Model-Based Recursive Self-Improvement through Reflective Rulebooks", "published": "2026-10-08T12:06:54Z", "decision": "reviewed_deferred", "reason": "冻结LLM生成可修订natural-language rulebook并编译可执行world model；预测误差驱动修订，接受必须同时LLM faithful判定与cell-exact历史transition replay；population共享观察探索。; 25 public ARC-AGI-3 games全部实际交互动作计费，不看environment source；Pong三开局episode闭环。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11794v1#S2", "https://arxiv.org/html/2610.11794v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11794v1", "sha256": "456fd9e557b2e5aa08e6eb155effbe17b27af2b8c0d70fa8dd91afff3d66e4f8" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11776", "title": "DPPM: Dual-Path Parametric Memory for Personalized Language Models", "published": "2026-10-08T11:56:07Z", "reviews": [ { "id": "2610.11776", "title": "DPPM: Dual-Path Parametric Memory for Personalized Language Models", "published": "2026-10-08T11:56:07Z", "decision": "reviewed_deferred", "reason": "DPPM由frozen compiler编码historysegment且排除query/answer；Evidence coordinatewise temporalsoftmax直接池化并RMS校准，Delta用normalizedELUkey、sigmoidgate对associativestate写预测残差并平行onesnormalizer。learnedfuse两路径，经frozen decoder生成historyconditionedLoRA，与sharedadapter相加；answerCE仅训练memorymodules。; PersonaMemv2 18527train2059val5000test且testpersona不入train/val；PrefEval seed42 16train/4heldouttopics，7380train1620test、10/70/300turn，2B-7Bbackbones；报告accuracy及Evidence/Delta/目标ablation，额外FKL/GRPO并非全部改善。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11776v1#S3", "https://arxiv.org/html/2610.11776v1#S4", "https://arxiv.org/html/2610.11776v1#Sx1" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11776v1", "sha256": "8ec004c5cd91a52e2559bc46ae9af6a32bc9a59bd777bbfe0e52336805077c89" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11773", "title": "Safe Actions Alone Do Not Ensure Safe Agents: Identifying Unfulfilled Obligations with Guard Models", "published": "2026-10-08T11:54:04Z", "reviews": [ { "id": "2610.11773", "title": "Safe Actions Alone Do Not Ensure Safe Agents: Identifying Unfulfilled Obligations with Guard Models", "published": "2026-10-08T11:54:04Z", "decision": "reviewed_deferred", "reason": "已读ObligationBench定义/实验：guard识别未履行required safety actions而不只forbidden，240专家trajectory和40k合成training。下一批要真实guard训练/完整set recall，不能用high task accuracy掩盖未履行义务。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "5. Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3. ObligationBench", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11773v1", "sha256": "c015aaac6b4a88b714b3dc35b4dd019a8b19f8484efc3d785ef4a8d8b415cf65", "local_html_sha256": "c015aaac6b4a88b714b3dc35b4dd019a8b19f8484efc3d785ef4a8d8b415cf65", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11773v1", "result": "retrieved", "sha256": "c015aaac6b4a88b714b3dc35b4dd019a8b19f8484efc3d785ef4a8d8b415cf65" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11768", "title": "Narrow and Deep: An Ontology Tower as the Knowledge of an LLM Agent for an Industrial Equipment System", "published": "2026-10-08T11:50:51Z", "reviews": [ { "id": "2610.11768", "title": "Narrow and Deep: An Ontology Tower as the Knowledge of an LLM Agent for an Industrial Equipment System", "published": "2026-10-08T11:50:51Z", "decision": "reviewed_deferred", "reason": "已读ontology tower方法/结果：物理derived quantities/operating lesson nodes projected给LLM，真实PLC plant九task preregistered replay及14live runs、invariant安全层。是真设备case证据但非随机生产A/B；下一批仅安全离线replay，不能据论文授权控制现实设备。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11768v1", "sha256": "469c70c1b90fe8df0faeb0aeacf2aec645d23583c4cb867487e2f4af86fe531c", "local_html_sha256": "469c70c1b90fe8df0faeb0aeacf2aec645d23583c4cb867487e2f4af86fe531c", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11768v1", "result": "retrieved", "sha256": "469c70c1b90fe8df0faeb0aeacf2aec645d23583c4cb867487e2f4af86fe531c" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11766", "title": "Same Outcome, Different Evidence: Intent Recovery in LLM Safety Evaluation", "published": "2026-10-08T11:50:32Z", "reviews": [ { "id": "2610.11766", "title": "Same Outcome, Different Evidence: Intent Recovery in LLM Safety Evaluation", "published": "2026-10-08T11:50:32Z", "decision": "reviewed_deferred", "reason": "IRIS将operative understandingrate UR与ASR并报；UR要求识别且把目标作为正在回答的任务，task-aware refusal/safealternative算recovered，纯翻译/语言分析不算。分开taskrecognition、responseframe、harmfulassistance、blocked/empty，后两仍计两denominator。; 105去重可跨接口有害任务、六closed/API模型，English direct与semanticcompressed接口及渐明确English reconstruction；similarASR对应不同UR，FormalLogic高UR也可高ASR。人类agreement/judge敏感度与bootstrap量化标签不确定性。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11766v1#S2", "https://arxiv.org/html/2610.11766v1#S3", "https://arxiv.org/html/2610.11766v1#S4", "https://arxiv.org/html/2610.11766v1#Sx1" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11766v1", "sha256": "b8f766d9c46819eeb94c9e9b93334f7a2af915c1faa4a82b84439ed38a7c2363" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11761", "title": "RAGenome: Scaling Retrieval-Based Genomic Language Models to Long Contexts", "published": "2026-10-08T11:44:58Z", "reviews": [ { "id": "2610.11761", "title": "RAGenome: Scaling Retrieval-Based Genomic Language Models to Long Contexts", "published": "2026-10-08T11:44:58Z", "decision": "reviewed_deferred", "reason": "RAGenome读取100vertebrate whole-genomealignment，丢gap却保留originalalignmentcolumn RoPE，按phylogenetic tree固定tokenbudget挑lineage，taxonomy各级共享embedding。独立species slidingwindowlocalattention+middlequery对homologcrossattention，clade masking MLM、分阶段扩展13312context、168M模型。; BEND genefindingfrozenembedding+two-layerCNN、官方test MCC约.45→.60；ClinVar zero-shot variantprioritization维持演化任务能力；最大retrievalbudget80000适配H100，较MSA方法减少gap训练token。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11761v1#S3", "https://arxiv.org/html/2610.11761v1#S4", "https://arxiv.org/html/2610.11761v1#S5", "https://arxiv.org/html/2610.11761v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11761v1", "sha256": "18f058ca100e3fe72cc7f1c5256022c4f71d9f0388988d75cfa9cf5b0475334f" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11732", "title": "MemTrial: Learning When to Trust Memory in LLM Portfolio Agents", "published": "2026-10-08T11:27:49Z", "reviews": [ { "id": "2610.11732", "title": "MemTrial: Learning When to Trust Memory in LLM Portfolio Agents", "published": "2026-10-08T11:27:49Z", "decision": "reviewed_deferred", "reason": "已读MemTrial方法/实验：同date八fractional-factorial memory组合draft估Banzhaf差值抵消market move，hierarchical Bayesian pool并在unseen dates预测成立才启用，保守referencefallback。下一批研究需时间隔离真实market return，不作投资建议或test未来信息输入。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4. Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3. MemTrial: Learning When to Trust Memory", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11732v1", "sha256": "46e98b2889806514fc5160741c0e0a5bc7106e30960a3e3b485fce8270b16b99", "local_html_sha256": "46e98b2889806514fc5160741c0e0a5bc7106e30960a3e3b485fce8270b16b99", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11732v1", "result": "retrieved", "sha256": "46e98b2889806514fc5160741c0e0a5bc7106e30960a3e3b485fce8270b16b99" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11725", "title": "Implementing the Spec Growth Engine: Preventing Spec-Code Divergence, and Growing the Spec with Agents", "published": "2026-10-08T11:24:55Z", "reviews": [ { "id": "2610.11725", "title": "Implementing the Spec Growth Engine: Preventing Spec-Code Divergence, and Growing the Spec with Agents", "published": "2026-10-08T11:24:55Z", "decision": "reviewed_deferred", "reason": "已读Spec Growth Engine两层/operation：deterministic spec/import graph、recorded test gate，与三独立model角色分离，18运行delegation开关。尚未在限定读段核验quantitative比较，保留工程候选但不将设计描述当能力实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 The Engine", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Operation: From by Hand to Unsupervised", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11725v1", "sha256": "28c68e1d2aaa4cc91cbba66744afc97cca3eab39d7ec1c1fe67ed8a1cce1b338", "local_html_sha256": "28c68e1d2aaa4cc91cbba66744afc97cca3eab39d7ec1c1fe67ed8a1cce1b338", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11725v1", "result": "retrieved", "sha256": "28c68e1d2aaa4cc91cbba66744afc97cca3eab39d7ec1c1fe67ed8a1cce1b338" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11716", "title": "4-Tensor Attention Model for Semantic Physical Reality", "published": "2026-10-08T11:19:23Z", "reviews": [ { "id": "2610.11716", "title": "4-Tensor Attention Model for Semantic Physical Reality", "published": "2026-10-08T11:19:23Z", "decision": "rejected_or_out_of_scope", "reason": "已读4-tensor设计/实验：中心是ROCStories nextsentence语义update网络，robot planner/encoder/renderer明确在外；单seed、baseline逐token free-running造成时间协议差异。不是本批agent执行算法，不把故事CE推广机器人能力。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 4-Tensor Transformer Design", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "8 Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11716v1", "sha256": "e7c35191c5196e743c93fcd81dce9d12b813bdd93b3edf085bf29e5a8a43d24e", "local_html_sha256": "e7c35191c5196e743c93fcd81dce9d12b813bdd93b3edf085bf29e5a8a43d24e", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11716v1", "result": "retrieved", "sha256": "e7c35191c5196e743c93fcd81dce9d12b813bdd93b3edf085bf29e5a8a43d24e" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.11696", "title": "A 3D Characterization Framework for Intelligent Sequential Decision Making", "published": "2026-10-08T11:06:51Z", "reviews": [ { "id": "2610.11696", "title": "A 3D Characterization Framework for Intelligent Sequential Decision Making", "published": "2026-10-08T11:06:51Z", "decision": "reviewed_deferred", "reason": "已读3D characterization与methods：MDP投影、人类设计prior和skill/cost比较graph/RL/LLM在Tower of Hanoi，包括DA-ToS。主要controlled puzzle methodology、并非通用agent能力，留待下一批评测协议核验不优先实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "III Proposed 3D Characterization Framework", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "IV Methods", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11696v1", "sha256": "41ca663ac2ce559150d13432bf7808711291d9401d560b1e4f40bd98090abd0a", "local_html_sha256": "41ca663ac2ce559150d13432bf7808711291d9401d560b1e4f40bd98090abd0a", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11696v1", "result": "retrieved", "sha256": "41ca663ac2ce559150d13432bf7808711291d9401d560b1e4f40bd98090abd0a" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11678", "title": "TRACE: Diagnosing Verifier Brittleness in Agentic Evaluation", "published": "2026-10-08T10:52:26Z", "reviews": [ { "id": "2610.11678", "title": "TRACE: Diagnosing Verifier Brittleness in Agentic Evaluation", "published": "2026-10-08T10:52:26Z", "decision": "reviewed_deferred", "reason": "TRACE区分行为变化与verifier变化，paired mutation和same-trajectory rescoring在tau2-bench有重复/positive controls；synthetic agents可读intended fix只属diagnostic，下一批转evaluation integrity，不当capability implementation。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "2 TRACE: From Score Differences to Explanations", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Study 2: LLM Agents on τ 2 \\tau^{2} -bench", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11678v1", "sha256": "b642aed3ab93ab210b67c2973285657af44253eb668eda4e4fb39f47d99aed86", "local_html_sha256": "b642aed3ab93ab210b67c2973285657af44253eb668eda4e4fb39f47d99aed86", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11678v1", "result": "retrieved", "sha256": "b642aed3ab93ab210b67c2973285657af44253eb668eda4e4fb39f47d99aed86" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11674", "title": "VESSI - VLM-Enhanced Support for Surveillance and Investigations", "published": "2026-10-08T10:50:17Z", "reviews": [ { "id": "2610.11674", "title": "VESSI - VLM-Enhanced Support for Surveillance and Investigations", "published": "2026-10-08T10:50:17Z", "decision": "rejected_or_out_of_scope", "reason": "VESSI按已知video类别给独立抽帧VLM category-specificprompt，response归Yes/No/Unclear，平均生成token最大softmax只作within-modelsalience筛选。CMUS组合coverage、相对视频时长processingtime、Yesrate及固定Yesratebucket惩罚，权重1/1/-.5/-1。; UCF-Crime950视频/13类、四约7Boff-the-shelf模型，manual65videos5909frames与50normalvideo补充排序检查；原文报告>66%video至少flag、>85%时间节约，但不是完整frameGT或真实人工reviewtime试验。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11674v1#S3", "https://arxiv.org/html/2610.11674v1#S4", "https://arxiv.org/html/2610.11674v1#S5", "https://arxiv.org/html/2610.11674v1#S6", "https://arxiv.org/html/2610.11674v1#S8" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11674v1", "sha256": "cdd522f9e350ccf9239e8e4d713bba393d38e8545c27ff94b29c57ff2ca0040f" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.11663", "title": "Constrained Command-Conditioned Reinforcement Learning with Bandit Strategy Selection in Real-Time Strategy Games", "published": "2026-10-08T10:37:33Z", "reviews": [ { "id": "2610.11663", "title": "Constrained Command-Conditioned Reinforcement Learning with Bandit Strategy Selection in Real-Time Strategy Games", "published": "2026-10-08T10:37:33Z", "decision": "rejected_or_out_of_scope", "reason": "已读method/results：MicroRTS受约束command PPO executor、Thompson sampling策略bandit，观察推断opponent而非身份；传统RL游戏agent非LLM，当前LLM研究批范围外。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11663v1", "sha256": "32e31ba7824f4fdcebd402519a65e4992b5dc9a4d4d8dcf40df1a03721a218ab", "local_html_sha256": "32e31ba7824f4fdcebd402519a65e4992b5dc9a4d4d8dcf40df1a03721a218ab", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11663v1", "result": "retrieved", "sha256": "32e31ba7824f4fdcebd402519a65e4992b5dc9a4d4d8dcf40df1a03721a218ab" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.11659", "title": "DIAL-OPD: Learning More from Fewer Tokens in On-Policy Distillation", "published": "2026-10-08T10:36:00Z", "reviews": [ { "id": "2610.11659", "title": "DIAL-OPD: Learning More from Fewer Tokens in On-Policy Distillation", "published": "2026-10-08T10:36:00Z", "decision": "accepted", "reason": "DIAL-OPD已有实现：logarithmic mean与density exponent调节token gap，retention40%多个teacher/student；本次核对论文选择机制与seven-benchmark口径，不新增实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "4 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "5 Experimental Setup", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11659v1", "sha256": "7e980643b14c22e3c39588e084182f74ad107b56ad6aa96333a9aaa5cb7ae3a9", "local_html_sha256": "7e980643b14c22e3c39588e084182f74ad107b56ad6aa96333a9aaa5cb7ae3a9", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11659v1", "result": "retrieved", "sha256": "7e980643b14c22e3c39588e084182f74ad107b56ad6aa96333a9aaa5cb7ae3a9" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "accepted" }, { "id": "2610.11655", "title": "Harness Evolution Hits a Ceiling: When Weight Training Should Begin", "published": "2026-10-08T10:33:29Z", "reviews": [ { "id": "2610.11655", "title": "Harness Evolution Hits a Ceiling: When Weight Training Should Begin", "published": "2026-10-08T10:33:29Z", "decision": "reviewed_deferred", "reason": "已读failure taxonomy与实验协议：harness/process vs content失误、2×2 evolved/base harness与trained/base LoRA以及answer-shuffle placebo；48dev/72heldout，自定harness替代officialloop须明确。下一批需实际权重训练和语义失败checker，不能promote诊断score为论文能力。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 A Diagnostic Framework", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11655v1", "sha256": "8e146ba9cfb15ddefceaa59286fbc0b5349ed8f79c1382991f0e1ac62881e0e8", "local_html_sha256": "8e146ba9cfb15ddefceaa59286fbc0b5349ed8f79c1382991f0e1ac62881e0e8", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11655v1", "result": "retrieved", "sha256": "8e146ba9cfb15ddefceaa59286fbc0b5349ed8f79c1382991f0e1ac62881e0e8" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11651", "title": "Evidence-Traceable Dynamic Interviewer Architecture for Expertise-Adaptive Qualitative Interviews Using Local LLMs", "published": "2026-10-08T10:26:13Z", "reviews": [ { "id": "2610.11651", "title": "Evidence-Traceable Dynamic Interviewer Architecture for Expertise-Adaptive Qualitative Interviews Using Local LLMs", "published": "2026-10-08T10:26:13Z", "decision": "reviewed_deferred", "reason": "已读local interviewer架构/实验：五prompt模块、持久interview state、expertise profiling与uniqueness，246真实参与者但主要一致性/问卷关联，无随机对照核验。保留nextbatch工程候选，participant满意度不当task capability/productionAB。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Experiment and Verification Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Evidence-Traceable Dynamic Interviewer Architecture", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11651v1", "sha256": "bad7a46be3bf42452ee8fd562a81f4bbfdbc229d786ea993ea834344020694dc", "local_html_sha256": "bad7a46be3bf42452ee8fd562a81f4bbfdbc229d786ea993ea834344020694dc", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11651v1", "result": "retrieved", "sha256": "bad7a46be3bf42452ee8fd562a81f4bbfdbc229d786ea993ea834344020694dc" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11647", "title": "One Skill Too Many: How Co-Installed Skills Conflict in Coding Agents", "published": "2026-10-08T10:20:47Z", "reviews": [ { "id": "2610.11647", "title": "One Skill Too Many: How Co-Installed Skills Conflict in Coding Agents", "published": "2026-10-08T10:20:47Z", "decision": "reviewed_deferred", "reason": "已读skill冲突study/results：20k repo矿、312确认pairs真实工具trace，exclusive core functions独立于task完成；firstread hook恢复规范 fidelity。下一批eval infra需冻结skill provenance与权限，不能仅test completion或把第三方SKILL当执行authority。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3. Study Design", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4. Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11647v1", "sha256": "172df5048ebd0a7aaa0106cde955a4eabb6a6efd63db0ae9784b5900d552ccdd", "local_html_sha256": "172df5048ebd0a7aaa0106cde955a4eabb6a6efd63db0ae9784b5900d552ccdd", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11647v1", "result": "retrieved", "sha256": "172df5048ebd0a7aaa0106cde955a4eabb6a6efd63db0ae9784b5900d552ccdd" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11638", "title": "UXBench Pro: Benchmarking Personalized User Experience in Multi-Turn Dialogue Interactions", "published": "2026-10-08T10:15:55Z", "reviews": [ { "id": "2610.11638", "title": "UXBench Pro: Benchmarking Personalized User Experience in Multi-Turn Dialogue Interactions", "published": "2026-10-08T10:15:55Z", "decision": "reviewed_deferred", "reason": "UXBench Pro真实交互test+FACTORS profiles，并URM/Sim4Eval及两meta-bench校验human alignment；private人口日志/用户profile要额外data协议，下一批evaluation/per-user reward候选，不与gold评分混用。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 UXBench Pro", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Meta-Benchmarking", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11638v1", "sha256": "0067ba18f5a076d9b70a088728fb4340ba0c6eeb1039769579ea5af911a04e60", "local_html_sha256": "0067ba18f5a076d9b70a088728fb4340ba0c6eeb1039769579ea5af911a04e60", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11638v1", "result": "retrieved", "sha256": "0067ba18f5a076d9b70a088728fb4340ba0c6eeb1039769579ea5af911a04e60" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11618", "title": "PolyCodeEval: Benchmarking Multilingual Code Generation from Functions to Repositories", "published": "2026-10-08T09:58:32Z", "reviews": [ { "id": "2610.11618", "title": "PolyCodeEval: Benchmarking Multilingual Code Generation from Functions to Repositories", "published": "2026-10-08T09:58:32Z", "decision": "reviewed_deferred", "reason": "已读PolyCodeEval设计/benchmark：58真实可执行repo、多语言/粒度2590task统一execution，same oracle pool控制domain。下一批需各granularity独立integration与hiddenoracle，不能用functionpass表示repo能力。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4. Case Study Design", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3. PolyCodeEval Benchmark", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11618v1", "sha256": "40fe59d8761df64f4ee54fa37bd6a45eac60dbe2351826d55c56107852ebce92", "local_html_sha256": "40fe59d8761df64f4ee54fa37bd6a45eac60dbe2351826d55c56107852ebce92", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11618v1", "result": "retrieved", "sha256": "40fe59d8761df64f4ee54fa37bd6a45eac60dbe2351826d55c56107852ebce92" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11610", "title": "Beyond Report Imitation: Clinically Aware Multi-Image Ultrasound Report Generation from Visible Evidence", "published": "2026-10-08T09:53:46Z", "reviews": [ { "id": "2610.11610", "title": "Beyond Report Imitation: Clinically Aware Multi-Image Ultrasound Report Generation from Visible Evidence", "published": "2026-10-08T09:53:46Z", "decision": "reviewed_deferred", "reason": "CAMEO将两张archivedultrasound原报告拆atomicpoint，以两个LLMverifier+arbiter分visible/partial/unsupported/excluded和supportview构造trustedtemplate，visible正例、partial qualified。StageIprojector对齐+singleimageadapt、StageIIcrossviewQA/reportSFT、StageIII临床错误chosen/rejected的DPO；visionencoder冻结，后两阶段更新projector/LLM。; PubMedVision-US instruction25631、USReportDistilled5890train/1474test及17670traininginstances、Pref21869train1354val不作为testreference。ClinicalScore55.02→74.20；blindreader301atomicannotationsteacheraccuracy78.9%/F184.8%，rawpoint25.2%unsupported。StageII主要lexicalgain、StageIII重塑errorprofile。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11610v1#S3", "https://arxiv.org/html/2610.11610v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11610v1", "sha256": "52d6342c78a32fe56783c96f832f00220732593c659c5a9dce95d49a94cfa459" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11605", "title": "NanoProof: Open and Efficient Automated Theorem Proving in Lean 4", "published": "2026-10-08T09:49:39Z", "reviews": [ { "id": "2610.11605", "title": "NanoProof: Open and Efficient Automated Theorem Proving in Lean 4", "published": "2026-10-08T09:49:39Z", "decision": "reviewed_deferred", "reason": "NanoProof factorizedAND-OR MCTS actor生成Lean tactic、critic预测negativeproofsubtreedepth，Lean立即执行丢invalidtactic，所有分支需闭合；matchmaker重加权能力边缘theorem，successfulproofreplay混SFT循环训练。LeanTree从人写proof拆结构tree、支持中间state重启与simplification/diversetacticsampling，最终导出standaloneLean独立验证。; 10B Nemotronmathbasetrain、Mathlib287kproofsteps文件最后10%29kval、258ktrain；MiniF2FValid手调参，Testfinal，三independentrunsmin/max。报告Test50.8%pass@16及约90x/7xclassbaselinecompute下降，hardware/FLOP部分由作者估算。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11605v1#S3", "https://arxiv.org/html/2610.11605v1#S4", "https://arxiv.org/html/2610.11605v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11605v1", "sha256": "454a500d095ec3205b70847267b9f4f111faa44aa2aa5554be08826e70c65bab" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11602", "title": "Where Do the Tokens Go? Understanding and Reducing Costs in LLM Agents for Vulnerability Discovery", "published": "2026-10-08T09:47:18Z", "reviews": [ { "id": "2610.11602", "title": "Where Do the Tokens Go? Understanding and Reducing Costs in LLM Agents for Vulnerability Discovery", "published": "2026-10-08T09:47:18Z", "decision": "reviewed_deferred", "reason": "已读CyberGym token study/AVRI evaluation：200trace open coding诊断，BET将source支持与hypothesis分开，另20task测fullrun成本/成功。下一批只在公开隔离benchmark授权操作，失败和helper开销计入，不按低token单独宣传效率。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2. Empirical Study Design", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5. Evaluation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11602v1", "sha256": "80dc9f336a01dadb6e9fbc827a29c6e4b0870626f85fb57c42b6579598a9c0f7", "local_html_sha256": "80dc9f336a01dadb6e9fbc827a29c6e4b0870626f85fb57c42b6579598a9c0f7", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11602v1", "result": "retrieved", "sha256": "80dc9f336a01dadb6e9fbc827a29c6e4b0870626f85fb57c42b6579598a9c0f7" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11600", "title": "Error-Propagation Modeling for Failure Attribution in LLM-Based Multi-Agent Systems", "published": "2026-10-08T09:44:59Z", "reviews": [ { "id": "2610.11600", "title": "Error-Propagation Modeling for Failure Attribution in LLM-Based Multi-Agent Systems", "published": "2026-10-08T09:44:59Z", "decision": "reviewed_deferred", "reason": "已读EMFA方法/experiment：failed MAS dependency graph、cascade与persistent loop，propagation候选筛选后counterfactual correction验证decisive agent-step。Who&When58+126logs；下一批不把annotated decisive gold步骤给被评attributor。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "Experiment Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11600v1", "sha256": "1c3b81a62209fc8ef5fac438a57e542bb324929e62055cef14ac2344c941fbd1", "local_html_sha256": "1c3b81a62209fc8ef5fac438a57e542bb324929e62055cef14ac2344c941fbd1", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11600v1", "result": "retrieved", "sha256": "1c3b81a62209fc8ef5fac438a57e542bb324929e62055cef14ac2344c941fbd1" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11593", "title": "Runnable Commit Untangling for Coding Agents", "published": "2026-10-08T09:42:27Z", "reviews": [ { "id": "2610.11593", "title": "Runnable Commit Untangling for Coding Agents", "published": "2026-10-08T09:42:27Z", "decision": "reviewed_deferred", "reason": "已读RucTangle/TangleEval方法/实验：真实apply每commit、runtimecheck并调整分组顺序，最终code恒同patch；131patch及453regression repair。下一批需执行每intermediateversion、修复评测独立，不只文本拆commit。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11593v1", "sha256": "6428c0b3b2e6f838b52ad6b652e3de9f55b549bd41269d29842d2e6cac43c9d5", "local_html_sha256": "6428c0b3b2e6f838b52ad6b652e3de9f55b549bd41269d29842d2e6cac43c9d5", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11593v1", "result": "retrieved", "sha256": "6428c0b3b2e6f838b52ad6b652e3de9f55b549bd41269d29842d2e6cac43c9d5" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11592", "title": "Probing for Long-Horizon Deductive Reasoning Capabilities in Language Models with Prolog", "published": "2026-10-08T09:42:07Z", "reviews": [ { "id": "2610.11592", "title": "Probing for Long-Horizon Deductive Reasoning Capabilities in Language Models with Prolog", "published": "2026-10-08T09:42:07Z", "decision": "reviewed_deferred", "reason": "ProloNg合成无loop/recursion/negation、arity≤2、eachpredicateone rule的definitePrologDAG；GenerateRule保证一条深Npath、Poissonbodybreadth，生成pairedprovable/unprovablequeries并mutation单fact、shuffle全部rule/fact。sharedcontext扩展同program不同depthquery隔离长度。; 8reasoningmodels、depth至22/62kcontext，多数depth>10近balancedchance.5；原depth/lengthr.97 GLMM无法可靠拆因果。另50sharedcontexts×6depth×正负共600samples，r.0003，四openmodels两因素均显著影响。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11592v1#S3", "https://arxiv.org/html/2610.11592v1#S4", "https://arxiv.org/html/2610.11592v1#Sx1" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11592v1", "sha256": "5833142aea2a4304c6b87492bb821de84bb7c28673671b8ef7f2bc19335d789f" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11585", "title": "Adapting English Quality Classifiers for Multilingual LLM Pretraining Data Selection", "published": "2026-10-08T09:33:32Z", "reviews": [ { "id": "2610.11585", "title": "Adapting English Quality Classifiers for Multilingual LLM Pretraining Data Selection", "published": "2026-10-08T09:33:32Z", "decision": "reviewed_deferred", "reason": "将FineWeb2 top100language-scriptpair原文0.1%样本Qwen3-32B译English，再EnglishFineWebedu/DCLM教师打分；frozenXLMR/mmBERT原语言embedding上two3072ReLUhiddenMLP回归，edu[0,5]L1、DCLMclipprob后logitL1。约6M样本包含1MEnglish/5Mmulti，top10%perlanguage过滤并保持原languagepretrainmix。; Apertus架构1B/3B/8B、4096context、AdEMAMix/WSD、约33kH100hourspretrain；译文约750H100hours。对比multilingualfilterbaseline、regional/culturalknowledge及未训练语言syntheticvswebteacher相关，报告相近downstream表现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11585v1#S3", "https://arxiv.org/html/2610.11585v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11585v1", "sha256": "1895e6c55e67ff63a8dad5b9e31c8bad2c984f13ae78f81ff77214eb721160a2" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11578", "title": "Chronos Enables Code Agents to Reason over Software Evolution", "published": "2026-10-08T09:31:37Z", "reviews": [ { "id": "2610.11578", "title": "Chronos Enables Code Agents to Reason over Software Evolution", "published": "2026-10-08T09:31:37Z", "decision": "reviewed_deferred", "reason": "已读Chronos方法/设置：历史PR蒸馏experience cards+typed graph、semantic entry+multi-hop、双patch生成steward选择；6backboneSWE对比。下一批须截止目标issue前history防答案PR泄漏，真实patch执行而非检索静态gold。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2. Approach", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3. Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11578v1", "sha256": "2bf3b9b1e522ffe508c1cb0f75c78494d7dd8bbb6c8df9417a6337df699b9975", "local_html_sha256": "2bf3b9b1e522ffe508c1cb0f75c78494d7dd8bbb6c8df9417a6337df699b9975", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11578v1", "result": "retrieved", "sha256": "2bf3b9b1e522ffe508c1cb0f75c78494d7dd8bbb6c8df9417a6337df699b9975" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11573", "title": "Memory Type Varies: Empowering LLM Agents for Long-Term Memory with Diverse Strategies", "published": "2026-10-08T09:28:31Z", "reviews": [ { "id": "2610.11573", "title": "Memory Type Varies: Empowering LLM Agents for Long-Term Memory with Diverse Strategies", "published": "2026-10-08T09:28:31Z", "decision": "reviewed_deferred", "reason": "已读MemoType/TriMEM方法和评测：learned router识别memory/query类型选择retrieval，Contriever统一backbone、LongMemEval/LoCoMo/PerLTQA。下一批须实际router训练与检索标签隔离，人工分类型heuristic不能称核心复现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Experiment", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 MemoType", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11573v1", "sha256": "2e47ed690d738d96365fa41bdcda378b0767a00b707d7bcee3c54faf0a3ab6c5", "local_html_sha256": "2e47ed690d738d96365fa41bdcda378b0767a00b707d7bcee3c54faf0a3ab6c5", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11573v1", "result": "retrieved", "sha256": "2e47ed690d738d96365fa41bdcda378b0767a00b707d7bcee3c54faf0a3ab6c5" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11561", "title": "Workerville: Towards an Organizational Behavior Account of Agent Safety", "published": "2026-10-08T09:23:28Z", "reviews": [ { "id": "2610.11561", "title": "Workerville: Towards an Organizational Behavior Account of Agent Safety", "published": "2026-10-08T09:23:28Z", "decision": "reviewed_deferred", "reason": "已读ACB formalization/Workerville：210task固定risk与16organizational antecedent配置，3360controlled challenges，人validatedjudge分别测披露/破坏/偏离。下一批安全infra需真实环境权限隔离，组织组合非单调结果不能泛化为现实企业因果。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Agentic Counterproductive Behavior", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Workerville", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11561v1", "sha256": "ef958d6085c4fef0ee7d6dd5e756a46e9d67160cf86fe91d1d48aa2bf3cf77af", "local_html_sha256": "ef958d6085c4fef0ee7d6dd5e756a46e9d67160cf86fe91d1d48aa2bf3cf77af", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11561v1", "result": "retrieved", "sha256": "ef958d6085c4fef0ee7d6dd5e756a46e9d67160cf86fe91d1d48aa2bf3cf77af" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11559", "title": "SWE-Journey: Towards More Realistic Evaluation of Coding Assistants through Long-Horizon, Multi-Turn Interaction", "published": "2026-10-08T09:21:49Z", "reviews": [ { "id": "2610.11559", "title": "SWE-Journey: Towards More Realistic Evaluation of Coding Assistants through Long-Horizon, Multi-Turn Interaction", "published": "2026-10-08T09:21:49Z", "decision": "reviewed_deferred", "reason": "已读SWE-Journey方法/实验：weak-to-strong task合成、固定F2P/P2P与persona user模拟真实多轮，13LLM mini-swe。下一批必须将完整requirement/goldpatch保留user/oracle侧，持续repo任务完成不能用单轮functionpass代理。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11559v1", "sha256": "50137ae09ce87af7aac113f23b38d70e4c1514bc4e50b6ab8a52213f2aac2319", "local_html_sha256": "50137ae09ce87af7aac113f23b38d70e4c1514bc4e50b6ab8a52213f2aac2319", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11559v1", "result": "retrieved", "sha256": "50137ae09ce87af7aac113f23b38d70e4c1514bc4e50b6ab8a52213f2aac2319" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11552", "title": "Safe, Persistent, and Evolving Agent Harness for Understanding Partially Observable Worlds", "published": "2026-10-08T09:15:28Z", "reviews": [ { "id": "2610.11552", "title": "Safe, Persistent, and Evolving Agent Harness for Understanding Partially Observable Worlds", "published": "2026-10-08T09:15:28Z", "decision": "reviewed_deferred", "reason": "已读E-Ledger/WorldAbduct方法/实验：everyaction policy gate、verified hiddenrule与dynamic state ledger，四视角abduction后targeted interaction验证才整合。WOW及科学环境安全completion/cost分开，下一批不让agent直接看完整隐藏数据库/worldtruth。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11552v1", "sha256": "f2b96c34f17c442f389c3f6638e72e8f6ee4689b9bc9f590beeb0907ae547484", "local_html_sha256": "f2b96c34f17c442f389c3f6638e72e8f6ee4689b9bc9f590beeb0907ae547484", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11552v1", "result": "retrieved", "sha256": "f2b96c34f17c442f389c3f6638e72e8f6ee4689b9bc9f590beeb0907ae547484" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11548", "title": "Conditional Transfer from Controlled Pretraining Mixtures to Code", "published": "2026-10-08T09:12:45Z", "reviews": [ { "id": "2610.11548", "title": "Conditional Transfer from Controlled Pretraining Mixtures to Code", "published": "2026-10-08T09:12:45Z", "decision": "reviewed_deferred", "reason": "固定70%CodeParrotPython+30%controlledsource simplex(OpenCodeInstruct/12curatedsynthetic/15literatureprobe)，固定模型/训练/Ft桥，分别定义diagnosticloss跟globalprogress、teachableloss跟ownanswer-tokenbudget、source对HumanEvaltransfer。Ado按近期loss拟合scalinglaw估未来1000step可下降量调source/taskprob。; 15coarsemixtures、60k/120kstepssparseexpertLM及固定1000updatecodeSFT；14/27tasksteachable但ranking不同于transfer。最优observedcontrolledslice.75OpenCode+.25curated pass20 22.6，纯Open19.5；三free-ratioAdo末Open仅1.2-1.4%、劣fixedcontrol2.4-11pp。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11548v1#S3", "https://arxiv.org/html/2610.11548v1#S4", "https://arxiv.org/html/2610.11548v1#S7", "https://arxiv.org/html/2610.11548v1#S9" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11548v1", "sha256": "d429903301ae1d95cc2673c9e79cd84f75618cce328db43b169447b4e813f44d" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11546", "title": "Learning to Orchestrate Evolutionary Search: Progression-Aware Deep Reinforcement Learning for Dynamic DE-CMA-ES Coordination in Optimization and Structural Model Updating", "published": "2026-10-08T09:12:13Z", "reviews": [ { "id": "2610.11546", "title": "Learning to Orchestrate Evolutionary Search: Progression-Aware Deep Reinforcement Learning for Dynamic DE-CMA-ES Coordination in Optimization and Structural Model Updating", "published": "2026-10-08T09:12:13Z", "decision": "rejected_or_out_of_scope", "reason": "已读DRL-DCO方法/results：DDPG动态协调DE/CMA-ES、population/diversity/restart用于结构FEMU与数值优化，非LLM/agentic research，本次范围外传统optimizer。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2. Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3. Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11546v1", "sha256": "d3dc3c1a912b82147dfeb249ad0a02e0e2213bb6febd8545446c89214e01c50a", "local_html_sha256": "d3dc3c1a912b82147dfeb249ad0a02e0e2213bb6febd8545446c89214e01c50a", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11546v1", "result": "retrieved", "sha256": "d3dc3c1a912b82147dfeb249ad0a02e0e2213bb6febd8545446c89214e01c50a" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.11543", "title": "Learning the Loop, Not Just the Page: Execution-Grounded Loop Learning for Web Generation", "published": "2026-10-08T09:11:17Z", "reviews": [ { "id": "2610.11543", "title": "Learning the Loop, Not Just the Page: Execution-Grounded Loop Learning for Web Generation", "published": "2026-10-08T09:11:17Z", "decision": "reviewed_deferred", "reason": "WebLoop共享Generator/Critic/Refiner通过browser execution rewards学习loop，critic先diagnostic后downstream helpfulness信用；WebRise/WebGen-Bench空workspace控制，下一批需真实网页执行与Critic训练。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 WebLoop: Learning the Generation–Critique–Refinement Loop", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11543v1", "sha256": "54e4531c7ca2ddb678f6a5a45e2bd24b369b59c7694352d1f97627ccba18e4a4", "local_html_sha256": "54e4531c7ca2ddb678f6a5a45e2bd24b369b59c7694352d1f97627ccba18e4a4", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11543v1", "result": "retrieved", "sha256": "54e4531c7ca2ddb678f6a5a45e2bd24b369b59c7694352d1f97627ccba18e4a4" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11542", "title": "Constitutional Gating and Deterministic Recovery for Multi-Agent LLM Negotiation: Ablations Against a Stateful Adversarial Gatekeeper", "published": "2026-10-08T09:10:45Z", "reviews": [ { "id": "2610.11542", "title": "Constitutional Gating and Deterministic Recovery for Multi-Agent LLM Negotiation: Ablations Against a Stateful Adversarial Gatekeeper", "published": "2026-10-08T09:10:45Z", "decision": "reviewed_deferred", "reason": "已读control-stack摘要、results/model-agnostic限制：Gatekeeper固定regex且已知canonical solution，purge+strike测deterministic recovery不是未知问题能力；5run/arm预注册token平均改善未支持。保留L1工程诊断，不计formal capability候选。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "6 Model-agnosticism and the\nedge", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11542v1", "sha256": "1b6c4ca12e936d59994b0d7c27118d886d1106bb9db43c955c060b12628b7a9a", "local_html_sha256": "1b6c4ca12e936d59994b0d7c27118d886d1106bb9db43c955c060b12628b7a9a", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11542v1", "result": "retrieved", "sha256": "1b6c4ca12e936d59994b0d7c27118d886d1106bb9db43c955c060b12628b7a9a" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11523", "title": "When to Intervene? State-Aware Sparse Manipulation in Federated Reinforcement Learning", "published": "2026-10-08T08:55:10Z", "reviews": [ { "id": "2610.11523", "title": "When to Intervene? State-Aware Sparse Manipulation in Federated Reinforcement Learning", "published": "2026-10-08T08:55:10Z", "decision": "rejected_or_out_of_scope", "reason": "已读V-BSA方法/实验：30-client Byzantine FRL中按policy uncertainty sparse poisoning，CartPole/Acrobot/LunarLander传统RL，不是LLM agent算法，本次范围外；不执行攻击代码。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11523v1", "sha256": "03ce8b3446efc3ed0be609f38b73cbd3b7102ad23804d1afd186fbaf0af0fcce", "local_html_sha256": "03ce8b3446efc3ed0be609f38b73cbd3b7102ad23804d1afd186fbaf0af0fcce", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11523v1", "result": "retrieved", "sha256": "03ce8b3446efc3ed0be609f38b73cbd3b7102ad23804d1afd186fbaf0af0fcce" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.11519", "title": "Residual Advantage: Student-Relative Teacher Guidance for RL with Verifiable Rewards", "published": "2026-10-08T08:52:55Z", "reviews": [ { "id": "2610.11519", "title": "Residual Advantage: Student-Relative Teacher Guidance for RL with Verifiable Rewards", "published": "2026-10-08T08:52:55Z", "decision": "accepted", "reason": "Residual Advantage/CoRA已有实现：bounded teacher-student probability residual减state value、response中心化保verifier mean，CoRA teacher LoRA逐轮更新；math三seed对照，保留existing实现不宣称新GPU证据。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "2 Residual Advantage", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "3 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11519v1", "sha256": "6c9bd85a439410f9ebce47dbb01f5f449ecc175878b03b55defd783eb10967de", "local_html_sha256": "6c9bd85a439410f9ebce47dbb01f5f449ecc175878b03b55defd783eb10967de", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11519v1", "result": "retrieved", "sha256": "6c9bd85a439410f9ebce47dbb01f5f449ecc175878b03b55defd783eb10967de" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "accepted" }, { "id": "2610.11514", "title": "SSCBench: Evaluating the Evidential Validity of Fault-Injection Tests for Tool-Using LLM Agents", "published": "2026-10-08T08:50:31Z", "reviews": [ { "id": "2610.11514", "title": "SSCBench: Evaluating the Evidential Validity of Fault-Injection Tests for Tool-Using LLM Agents", "published": "2026-10-08T08:50:31Z", "decision": "reviewed_deferred", "reason": "已读SSCBench协议/study：authority counterevidence可见时序需在firstuse前才支持相应claim，1191faulted tau-bench执行区分arrival、adoption与first faulty reliance。下一批重要评测infra，自动trace oracle须校准且不能把晚反证标成早期可用。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "V Empirical Study Design", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "II Fault-Injection Evaluation of Tool-Using Agents", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11514v1", "sha256": "fdfd6465c0aa4007a14b92c18226d250290ffad8d9bb7aa1742e045fb6a1cf79", "local_html_sha256": "fdfd6465c0aa4007a14b92c18226d250290ffad8d9bb7aa1742e045fb6a1cf79", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11514v1", "result": "retrieved", "sha256": "fdfd6465c0aa4007a14b92c18226d250290ffad8d9bb7aa1742e045fb6a1cf79" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11510", "title": "Does Modern Standard Arabic (MSA) Dominate Arabic Dialects in LLMs? A Representation-Level Analysis", "published": "2026-10-08T08:46:47Z", "reviews": [ { "id": "2610.11510", "title": "Does Modern Standard Arabic (MSA) Dominate Arabic Dialects in LLMs? A Representation-Level Analysis", "published": "2026-10-08T08:46:47Z", "decision": "reviewed_deferred", "reason": "在MADAR26Arabicvarieties语义parallel句抽各层hidden，NMI测dialectlabel/clusterseparability；用true-dialect与每个competitor alignedtokenposterior之比Lambda，汇总所有token/competitorpair中Lambda<1比例，不聚合单token竞争方向、不把低NMI等价MSAdominance。; BLOOM/mGPT每variety400句、附加model100句，Arabic平均NMI约.1/.2但confusion仍diagonal、linearprobe可预测；noevidenceMSAdominance，architecture separabilitypeak可后移，POS与首tokenablation。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11510v1#S3", "https://arxiv.org/html/2610.11510v1#S4", "https://arxiv.org/html/2610.11510v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11510v1", "sha256": "aa20200cb7b547b1b09271393e2af0ae2822bd3a44f81051be64763bbbf11ccd" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11508", "title": "WARP-VLA: Wrist-Camera Adaptation for View-Robust Policy Execution in Vision-Language-Action Models", "published": "2026-10-08T08:44:26Z", "reviews": [ { "id": "2610.11508", "title": "WARP-VLA: Wrist-Camera Adaptation for View-Robust Policy Execution in Vision-Language-Action Models", "published": "2026-10-08T08:44:26Z", "decision": "reviewed_deferred", "reason": "WARP-VLA冻结visionencoder/actionpolicy，在wristtoken接口MoE恢复canonicalfeatures；训练canonicalmeananchor作KV、perturbedtokens作Q的sharedcrossattentiontransformer加15expertLoRA。router用residual/anchormean,std,x/yspatialmoments，13translation+2tilt组内softmax各占1/2；每expertfeatureMSE及基于routing×exp(-error)×poseprior的stopgradresponsibility监督router。; 2000LIBEROdemonstrations40tasks重render10perturbations，训练8RTXPRO6000约30h；LIBERO/pi.5平均39.2→78.3%，LIBEROPlus也改善。实xArm6十task/每mount5trials，basepolicy在500realcanonicaldemo微调，adapter固定simulationtrained但重新计算realcanonicalanchor。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11508v1#S3", "https://arxiv.org/html/2610.11508v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11508v1", "sha256": "3b79d62544015202c5b690f973cd0af18985e006ae8e3b24c19a4f2e89b96ef7" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11501", "title": "Beyond Sequences: Distilling Structured Decision Memory for LLM Recommendation", "published": "2026-10-08T08:40:35Z", "reviews": [ { "id": "2610.11501", "title": "Beyond Sequences: Distilling Structured Decision Memory for LLM Recommendation", "published": "2026-10-08T08:40:35Z", "decision": "rejected_or_out_of_scope", "reason": "阿里MARI的PHDD/SDM记忆与检索LLM结构相关，但§V是私有MBR/PHDD离线NIP/DCP评测；正文online指推理/部署范式，没有本方法量化生产A/B或经批准的全流量证据。不能因工业署名或online inference词接受。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "bounded_methods_and_evaluation", "review_artifact": "research-audits/oct12-review-closure/recommendation-reviews.json", "evidence_locations": [ "§IV-A–IV-D", "§V Experiments Setup", "§VI Evaluation" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11501v1", "sha256": "306f5a312384b60c2a0182b79c986c1935ddafe1bcdeeb3b1e190c861b19bda2" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "recommendation" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.11482", "title": "Evaluating Local Language Model Agents for Reproducible Data Engineering: An Empirical Software Engineering Study of Mobility Workflows", "published": "2026-10-08T08:26:46Z", "reviews": [ { "id": "2610.11482", "title": "Evaluating Local Language Model Agents for Reproducible Data Engineering: An Empirical Software Engineering Study of Mobility Workflows", "published": "2026-10-08T08:26:46Z", "decision": "reviewed_deferred", "reason": "15 mobility artifacts任务、bounded workspace，deterministic parsing/execution/schema/domain checker；one-shot与tools闭环配置对照及失败日志。; 10本地Qwen配置×2模式×15任务×5重复，共1500 attempt，RTX4090；Q4与部分BF16。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11482v1#S3", "https://arxiv.org/html/2610.11482v1#S4", "https://arxiv.org/html/2610.11482v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11482v1", "sha256": "951965a04e196a8fce7214e8671c335b2e67677f0571dee4ba8ea510a3194d23" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11469", "title": "SAGE: Sink-Aware Guided Emphasis for Visual Grounding in Vision-Language Decoders", "published": "2026-10-08T08:18:58Z", "reviews": [ { "id": "2610.11469", "title": "SAGE: Sink-Aware Guided Emphasis for Visual Grounding in Vision-Language Decoders", "published": "2026-10-08T08:18:58Z", "decision": "reviewed_deferred", "reason": "SAGE对nativeattentionpre-softmax视觉keylogits在tokenalignedROI内+alpha5、ROI外-beta2；CLIPpatch-textcosine可queryconditioned，ViTrollout/DINOv3CLSattention是queryagnostic saliency，topk nearest-neighbor映射visualgrid。singlelayersweep诊断PIS与spectralconcentration，可做topklayers。; LLaVA1.5/Qwen2VL/InternVL3/NVILA多bench；主表每backbone/bench/ROIsource取最佳sweeplayer是upperbound。另LLaVA7B先按label-freeSinkIoU/top1spectralminimum选固定L24，再所有bench统一评估；TextVQA/GQA收益较大，其他小，same-size randomROI无收益。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11469v1#S3", "https://arxiv.org/html/2610.11469v1#S4", "https://arxiv.org/html/2610.11469v1#Sx1" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11469v1", "sha256": "bf4d760560bd39616145af718dc41d0f43c978074cd233803c5406784a79218b" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11464", "title": "Who Verifies the Verifier? Co-Evolving Inspectable Graders with Self-Improving Agents", "published": "2026-10-08T08:13:53Z", "reviews": [ { "id": "2610.11464", "title": "Who Verifies the Verifier? Co-Evolving Inspectable Graders with Self-Improving Agents", "published": "2026-10-08T08:13:53Z", "decision": "reviewed_deferred", "reason": "演化inspectable pass/fail/abstain grader以10-item dev anchors选择，test locked，MBPP+/Spider2/report workloads；适合verifier/skill co-evolution下一批，必须验证grader validity，downstream task lift不能替代。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 The verifier and its evolution", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experimental setup", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11464v1", "sha256": "db45d964d77020219f78316da71149f8a3bf31157e5fb69f7f96dc75679e2302", "local_html_sha256": "db45d964d77020219f78316da71149f8a3bf31157e5fb69f7f96dc75679e2302", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11464v1", "result": "retrieved", "sha256": "db45d964d77020219f78316da71149f8a3bf31157e5fb69f7f96dc75679e2302" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11461", "title": "Beyond Speech Captions: Speech-Rewarded Style Planning for Conversational Text-to-Speech", "published": "2026-10-08T08:12:28Z", "reviews": [ { "id": "2610.11461", "title": "Beyond Speech Captions: Speech-Rewarded Style Planning for Conversational Text-to-Speech", "published": "2026-10-08T08:12:28Z", "decision": "reviewed_deferred", "reason": "SRSP冻结TTS，以teacher-forced target speech-token likelihood作GRPO style-planner训练reward，推理只dialogue/text；scene/movie-disjoint dev/test划分明确，下一批需真实声学模型与语音评测。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Speech-Rewarded Style Planning", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11461v1", "sha256": "4d82b63ea2353f5f8d7151e291b4730899b3d373d8e8f20e52ef48498b683050", "local_html_sha256": "4d82b63ea2353f5f8d7151e291b4730899b3d373d8e8f20e52ef48498b683050", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11461v1", "result": "retrieved", "sha256": "4d82b63ea2353f5f8d7151e291b4730899b3d373d8e8f20e52ef48498b683050" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11454", "title": "Zatom-2: Multitask Pretraining on Atomistic Data for Generative Modeling across Domains", "published": "2026-10-08T08:09:17Z", "reviews": [ { "id": "2610.11454", "title": "Zatom-2: Multitask Pretraining on Atomistic Data for Generative Modeling across Domains", "published": "2026-10-08T08:09:17Z", "decision": "reviewed_deferred", "reason": "Zatom2 atom-token-atom U型localencoder/globaltokentrunk/localdecoder，molecule/materialatom1与proteinresidueatom14共享conditionalflowmatchinggeometry；time-gatedelement/sequence/fractional/latticeheads、optionalenergy/forceheads，多任务与跨OMol/OMat预训练，globalmeanforcenorm条件控制relaxationregime。; OMol25 3,986,754+OMat24 1,009,850约5M structures，80epochs多配置；UMA-S1p2计算生成样本forceconsistencyAUROC/MAE、分布fidelity，公开molecule/materialbench；2000protein-domainfinetune lengthextrapolationdesignability67.8→74.8%。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11454v1#S2", "https://arxiv.org/html/2610.11454v1#S3", "https://arxiv.org/html/2610.11454v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11454v1", "sha256": "e8bc1a9415480e12cceb250cbed5c9235776730619c2f65a6b84cee02285b4b2" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11451", "title": "SAIL: Scientific Agentic Intelligence via a Science-Aware Loop", "published": "2026-10-08T08:04:22Z", "reviews": [ { "id": "2610.11451", "title": "SAIL: Scientific Agentic Intelligence via a Science-Aware Loop", "published": "2026-10-08T08:04:22Z", "decision": "reviewed_deferred", "reason": "SAIL science-aware failure diagnosis/data construction循环，SFT→specialists→MOPD→agentic RL；AstaBench/SciCode/DRB2评测，science benchmarks与development隔离。需真实任务环境及35B backbone，冻结批次外待下一批。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "2 Science-Aware Improvement Loop", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Training Recipe", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Evaluation and Analysis", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11451v1", "sha256": "d9e33a3b9f1ad0b2d1d9271aef70de82ca6a4c50d74db115034d8d434230538b", "local_html_sha256": "d9e33a3b9f1ad0b2d1d9271aef70de82ca6a4c50d74db115034d8d434230538b", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11451v1", "result": "retrieved", "sha256": "d9e33a3b9f1ad0b2d1d9271aef70de82ca6a4c50d74db115034d8d434230538b" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11447", "title": "Closed-loop evaluation of LLM agents for embedded software development", "published": "2026-10-08T08:02:10Z", "reviews": [ { "id": "2610.11447", "title": "Closed-loop evaluation of LLM agents for embedded software development", "published": "2026-10-08T08:02:10Z", "decision": "reviewed_deferred", "reason": "plain-text firmware工程handoff+QEMU ESP32 constrained workspace，在one-shot/self-verify/CI red-green/oracle四种可见反馈下同pi agent wrapper执行build/runtime修复。; 7模型×5任务×4反馈×3重复420runs，同RTX4090/llama.cpp/GGUF环境仅local latency可比。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11447v1#S3", "https://arxiv.org/html/2610.11447v1#S4", "https://arxiv.org/html/2610.11447v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11447v1", "sha256": "b3aec43cd6a8e63e02c3c258c9a13b3cdae27665629fdb6eef8fbb69a664a252" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11430", "title": "BioBigBird: A Sparse Attention Model for Long-Range Dependency Processing in Biomedical Text", "published": "2026-10-08T07:54:12Z", "reviews": [ { "id": "2610.11430", "title": "BioBigBird: A Sparse Attention Model for Long-Range Dependency Processing in Biomedical Text", "published": "2026-10-08T07:54:12Z", "decision": "reviewed_deferred", "reason": "BioBigBird沿用BigBird稀疏双向4096context，PubMed intro/conclusion去重过滤后多阶段4096fullarticle→1024MIMICIII+abstract→512abstract预训练；NER tokenhead与RE pooledhead共享encoder、分别跨datasets CE multitaskfinetune。; PubMedarticle2.3M/10Btokens、abstract42M/10B、MIMIC2M/2B；三阶段1.89M/1.37M/1.5Msteps。BLURB13tasks与额外NER/RE，Stage3/MTStage3对PubMedBERT/BioLinkBERT/BERN2；并非每task都好，JNLPBA multitask78.55→77.40，QA多项较弱。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11430v1#S3", "https://arxiv.org/html/2610.11430v1#S4", "https://arxiv.org/html/2610.11430v1#S5", "https://arxiv.org/html/2610.11430v1#Sx1" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11430v1", "sha256": "49636dbd8c15d9b55e6548996c05590df4ed7049362bc743e011ccbe28af7b36" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11422", "title": "Causal-fate dynamics of unrealized influence", "published": "2026-10-08T07:48:53Z", "reviews": [ { "id": "2610.11422", "title": "Causal-fate dynamics of unrealized influence", "published": "2026-10-08T07:48:53Z", "decision": "reviewed_deferred", "reason": "causal-fate在realizedH、latentxi、completeZ=H+xi上按fullmapF和tokenlocalmapL构造h=F(Z)-L(H)，threshold每token释放整个h或保latent；Hnext=L(H)+psi，xinext=h-psi保证complete层递推严格fullmap。最终logit只realizedstate，terminallatent不强制释放。另connectome模拟与crossobserverBGPhistory现象为不同证据等级。; frozenSmolLM2-135M/WikiText2 length128，33valblocks4191positions/65test8255positions，validation选非trivial最低realizationratio且PPL≤1.06base；testfull38.6512/selective39.9138、ratio.884022。BGP publicRouteViewsChile/RIPErrc06 2025-07-06一天MRT、5,568,057records无parseerror；这是operationalnetworkobservationaldata，不是介入用户AB。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11422v1#S2", "https://arxiv.org/html/2610.11422v1#S3", "https://arxiv.org/html/2610.11422v1#S4", "https://arxiv.org/html/2610.11422v1#Sx1", "https://arxiv.org/html/2610.11422v1#Sx2", "https://arxiv.org/html/2610.11422v1#as1_Sx3", "https://arxiv.org/html/2610.11422v1#as1_Sx4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11422v1", "sha256": "21e8ea4a1b8053d415589bbd52464be1b52760ca6d00d3449626ab9940ea4c25" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11389", "title": "Fact over Fiction: Detection of Pathological Hallucinations in Sinhala-to-English Neural Machine Translation", "published": "2026-10-08T07:19:51Z", "reviews": [ { "id": "2610.11389", "title": "Fact over Fiction: Detection of Pathological Hallucinations in Sinhala-to-English Neural Machine Translation", "published": "2026-10-08T07:19:51Z", "decision": "reviewed_deferred", "reason": "SinhalaEnglishref-free detector mDeBERTa(S,H)tokenbinarylabel；从Englishreference做entity/antonym/dependency/numeric及EnglishreencodehighT的probabilisticchain，difflib标签replace/insert、charrescue.80处理形态变体，source/specialmask。neuralrisk、wordlogprob、LaBSE三signalensemble。; 24996cleanparallel抽7500source形成45000rows；detectorbalanced15000rows source-disjoint10486train1497val3017test/1485testsource。三seedF1.841±.001，shuffle/removesourceAUROC .970→chance，但hypothesisonlyF1.622；8NMTbenchmark仅detectorfiring51..433不是验证真实hallucinationrate。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11389v1#S3", "https://arxiv.org/html/2610.11389v1#S4", "https://arxiv.org/html/2610.11389v1#Sx1" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11389v1", "sha256": "5d745a06662a7fa1fe83ff07e7f818b19c69224902f3befcc005901f518188ec" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11384", "title": "Environmental Feedback Modeling Matters: Rethinking Feedback Treatment in Agentic Hindsight Self-Distillation", "published": "2026-10-08T07:16:00Z", "reviews": [ { "id": "2610.11384", "title": "Environmental Feedback Modeling Matters: Rethinking Feedback Treatment in Agentic Hindsight Self-Distillation", "published": "2026-10-08T07:16:00Z", "decision": "reviewed_deferred", "reason": "已读SELF方法/实验：on-policy真实environment反馈建feedback-conditioned hindsight teacher，joint response modeling loss+selfdistillation；tau/AppWorld reward-free agent训练。下一批须真实双objective梯度，privileged未来feedback仅训练teacher不进部署policy。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11384v1", "sha256": "2bf863a9273491625e0c73be16f7a74328331de538cf55ae27b897d25508f33d", "local_html_sha256": "2bf863a9273491625e0c73be16f7a74328331de538cf55ae27b897d25508f33d", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11384v1", "result": "retrieved", "sha256": "2bf863a9273491625e0c73be16f7a74328331de538cf55ae27b897d25508f33d" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11374", "title": "Rethinking Contrastive Loss in CLIP Post-training: A Complementary Framework with Frozen Text Encoder", "published": "2026-10-08T07:06:58Z", "reviews": [ { "id": "2610.11374", "title": "Rethinking Contrastive Loss in CLIP Post-training: A Complementary Framework with Frozen Text Encoder", "published": "2026-10-08T07:06:58Z", "decision": "reviewed_deferred", "reason": "ComCLIP冻结textencoder与DINOv2teacher，只更新CLIPvision单epoch：小tau.01 symmetricInfoNCE集中hardnegatives，postprojectionMSE锚originalCLIP，rowsoftmax selfsimilarity(T.1)的DINO→CLIP KL relationaldistill，不直接把CLIPfeature对齐DINOcoordinate。; ViTB16/L14、CC3M/12M/COCO caption、多seed；zero-shot≈CLIPRefine、linearprobeB16 48.99vs42.28、MMVPL14 24.20vs19.01，但retrievalCLIPRefine更强；dropinLLaVA1.5-7B projector/LLM不重align，8benchno netchange。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11374v1#S3", "https://arxiv.org/html/2610.11374v1#S4", "https://arxiv.org/html/2610.11374v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11374v1", "sha256": "3031139a15f19d315ec4b55a4bd1af636aa97ab1c9301f67a0de6f1e6f048fd1" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11373", "title": "From a Prompt to Repertoires: Evolving Functional REpertoires Enable LLM Continual Learning", "published": "2026-10-08T07:06:52Z", "reviews": [ { "id": "2610.11373", "title": "From a Prompt to Repertoires: Evolving Functional REpertoires Enable LLM Continual Learning", "published": "2026-10-08T07:06:52Z", "decision": "reviewed_deferred", "reason": "EFRE用历史一致性检测兼容更新或新增函数，在固定模型下external repertoire适应ToolUse→FinQA→SciKE-Bio；可作为agent continual learning候选，非权重后训练，下一批需要实际函数执行与遗忘评测。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11373v1", "sha256": "e4dd3319b80efaf58e3a7e9422c3495991171f875e201de48e9122965aafa5ef", "local_html_sha256": "e4dd3319b80efaf58e3a7e9422c3495991171f875e201de48e9122965aafa5ef", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11373v1", "result": "retrieved", "sha256": "e4dd3319b80efaf58e3a7e9422c3495991171f875e201de48e9122965aafa5ef" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11371", "title": "SignRAG: Unified Retrieval-Augmented Gloss-Free Sign Language Translation", "published": "2026-10-08T07:03:24Z", "reviews": [ { "id": "2610.11371", "title": "SignRAG: Unified Retrieval-Augmented Gloss-Free Sign Language Translation", "published": "2026-10-08T07:03:24Z", "decision": "reviewed_deferred", "reason": "SignRAG hierarchical sign/LLM alignment + train-only retrieval gallery + quality/utility rewards GRPO，四SLT公开split评测；下一批需实际sign encoder、retrieval与utility变化，不能gold translation进test检索。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiment", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11371v1", "sha256": "4b2b208b797987fbb34f71f87aef62c307f6bce1ba1c27f44b5bc5e2c73bf618", "local_html_sha256": "4b2b208b797987fbb34f71f87aef62c307f6bce1ba1c27f44b5bc5e2c73bf618", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11371v1", "result": "retrieved", "sha256": "4b2b208b797987fbb34f71f87aef62c307f6bce1ba1c27f44b5bc5e2c73bf618" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11370", "title": "RIT-RAG: Navigating Document Corpora with Retrieval-Induced Trees", "published": "2026-10-08T07:03:17Z", "reviews": [ { "id": "2610.11370", "title": "RIT-RAG: Navigating Document Corpora with Retrieval-Induced Trees", "published": "2026-10-08T07:03:17Z", "decision": "reviewed_deferred", "reason": "已读RIT-RAG corpus tree prelim/实验：retrieval候选位置诱导跨document可管理subtree，LLM选择读node/改query，2.84m webpageEntQA。下一批需实际retriever/tree navigation，不能静态gold章节选择或把live corpus大小当生产A/B。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "5 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Preliminaries", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11370v1", "sha256": "167b307c739efb9e740e99d6388a77e748ea2d521d814a10a1d7209349126f6d", "local_html_sha256": "167b307c739efb9e740e99d6388a77e748ea2d521d814a10a1d7209349126f6d", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11370v1", "result": "retrieved", "sha256": "167b307c739efb9e740e99d6388a77e748ea2d521d814a10a1d7209349126f6d" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11366", "title": "SpatialOPSD: Self-Distilling Spatial Intelligence from Verified Coding Agent Traces", "published": "2026-10-08T06:58:30Z", "reviews": [ { "id": "2610.11366", "title": "SpatialOPSD: Self-Distilling Spatial Intelligence from Verified Coding Agent Traces", "published": "2026-10-08T06:58:30Z", "decision": "reviewed_deferred", "reason": "SpatialOPSD用verified coding-agent traces为privileged teacher生成spatialCoT，repetition-aware mask+unlikelihood控leakage；5008原训练题student只见原input，SFT用不同4459数量应记录，不把等protocol称等data预算；下一批。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Spatial Reasoning via OPSD", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11366v1", "sha256": "3c5c06be29a5c47463999447782e96826ec01e077f3094939f6deae50733410e", "local_html_sha256": "3c5c06be29a5c47463999447782e96826ec01e077f3094939f6deae50733410e", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11366v1", "result": "retrieved", "sha256": "3c5c06be29a5c47463999447782e96826ec01e077f3094939f6deae50733410e" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11363", "title": "UniData: Universal Multimodal Instruction Generation Pipeline", "published": "2026-10-08T06:52:24Z", "reviews": [ { "id": "2610.11363", "title": "UniData: Universal Multimodal Instruction Generation Pipeline", "published": "2026-10-08T06:52:24Z", "decision": "reviewed_deferred", "reason": "UniData以keyword+随机context DEG生成多events，MIG textcenteredLLaMA3将Ovisvisual/Whispermusicembedding及其他六模态textrepresentation整合，预测image/musicprompt而非nativepixel/audio，调用SD3/MusicGen/emoji/search/qrcode实现输出。CIF窗口比较新旧round重复则提高repetitionpenalty、不相关则提高temperature，规则改sampling。; UniDataset20000entries/九输出类型，随机90%train10%test、冻结visual/musictokenizer；SelfInstructGPT4/VIGCVicuna基线与GPT-guidedquality、附加userstudy/downstreamfinetune及DEG/CIFablation。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11363v1#S3", "https://arxiv.org/html/2610.11363v1#S4", "https://arxiv.org/html/2610.11363v1#S5", "https://arxiv.org/html/2610.11363v1#Sx1" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11363v1", "sha256": "492f8cc1c6599112a8c11312d62c2f8be7e0576a2e62e4211b0df56e1e2552e3" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11361", "title": "SepGen: Multi-Stem Audio-Video Separation and Generation in a Single Model", "published": "2026-10-08T06:51:59Z", "reviews": [ { "id": "2610.11361", "title": "SepGen: Multi-Stem Audio-Video Separation and Generation in a Single Model", "published": "2026-10-08T06:51:59Z", "decision": "reviewed_deferred", "reason": "LTX2.5 audio stream增加两stem，mix/video不得attend stems且LoRA只在stem tokens启用；同timeline，caption block routing，separation→generation两阶段noise schedule与estimated-clean-mix inference。; §4涵盖194生成场景、Veo/LTX外部场景、172真实录音、三/五seed、swap margins与听觉人类对比。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract/author information/main defining methods/main experiments and tables; production-evidence paragraph spot check; appendices not exhaustively read", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11361v1#S3", "https://arxiv.org/html/2610.11361v1#S4", "https://arxiv.org/html/2610.11361v1#S4.SS4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11361v1", "sha256": "601890f045905d7abd7d526aa1074d34e8411720df458cfb737052967baadde5" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11358", "title": "RaReCache: Bridging the Gap in Cross-Model KV Cache Reuse via Rank disagreement-based Selective Recomputation", "published": "2026-10-08T06:49:28Z", "reviews": [ { "id": "2610.11358", "title": "RaReCache: Bridging the Gap in Cross-Model KV Cache Reuse via Rank disagreement-based Selective Recomputation", "published": "2026-10-08T06:49:28Z", "decision": "reviewed_deferred", "reason": "已读RaReCache方法/动机：真实linear KVmapping、calibration rank disagreement选token target重算，同familytokenizer身份必要；benchmarkaccuracy与singleGPU服务负载TTFT。下一批foundation/serving候选需真实异模型cache与CUDA，synthetic online serving不证明生产AB。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 Motivational Case Study", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11358v1", "sha256": "661416ef0466d2b4c68b5487ef9d1610f62133750843ce3c678a638b9e38e7a4", "local_html_sha256": "661416ef0466d2b4c68b5487ef9d1610f62133750843ce3c678a638b9e38e7a4", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11358v1", "result": "retrieved", "sha256": "661416ef0466d2b4c68b5487ef9d1610f62133750843ce3c678a638b9e38e7a4" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11354", "title": "AdaptEvo: Adaptive Agent Learning with Evolving Supervision", "published": "2026-10-08T06:47:41Z", "reviews": [ { "id": "2610.11354", "title": "AdaptEvo: Adaptive Agent Learning with Evolving Supervision", "published": "2026-10-08T06:47:41Z", "decision": "reviewed_deferred", "reason": "AdaptEvo CA-GRPO按reference confidence混合outcome/process，并演化decision knowledge/rubric；工业多模态moderation In/Out-Period评测，private data/changed rules使完整复现受限，下一批需公开替代语义与真实rubric闭环。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "4 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11354v1", "sha256": "27419bf408cb2407fedc5c74d59dd28ec0dd9c75a35c3c9a0a1f808b55476a12", "local_html_sha256": "27419bf408cb2407fedc5c74d59dd28ec0dd9c75a35c3c9a0a1f808b55476a12", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11354v1", "result": "retrieved", "sha256": "27419bf408cb2407fedc5c74d59dd28ec0dd9c75a35c3c9a0a1f808b55476a12" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11352", "title": "RL-ARC: Calibrating Large Reasoning Models via Reasoning-guided Uncertainty", "published": "2026-10-08T06:45:24Z", "reviews": [ { "id": "2610.11352", "title": "RL-ARC: Calibrating Large Reasoning Models via Reasoning-guided Uncertainty", "published": "2026-10-08T06:45:24Z", "decision": "reviewed_deferred", "reason": "RL-ARC以reasoning/answer verbalized confidences对correct regularize/incorrect overconfident penalize，在ID/OOD比较calibration与accuracy；下一批需proper scoring+真实reasoningconfidence，不能硬编码置信值。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 RL-ARC: Reinforcement Learning with Auxiliary Reasoning Confidence", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11352v1", "sha256": "0c105e996bf31fe0ddcdbbf8daee3426e630b0e92b70f50a66f866862b717507", "local_html_sha256": "0c105e996bf31fe0ddcdbbf8daee3426e630b0e92b70f50a66f866862b717507", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11352v1", "result": "retrieved", "sha256": "0c105e996bf31fe0ddcdbbf8daee3426e630b0e92b70f50a66f866862b717507" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11351", "title": "Deception by Omission: Language Models Knowingly Hide Their Mistakes", "published": "2026-10-08T06:45:02Z", "reviews": [ { "id": "2610.11351", "title": "Deception by Omission: Language Models Knowingly Hide Their Mistakes", "published": "2026-10-08T06:45:02Z", "decision": "reviewed_deferred", "reason": "已读prefilled mistake方法/结果：360chat和49agent六模拟环境，分aware conceal与未awareness及outside reviewer。下一批安全行为评测需标注prefill合成而非真实部署发生率，思维记录不是所有模型意识groundtruth。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Methods", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11351v1", "sha256": "01d0b477ce4b1567e5d2684a9d591a756b5a32c26a577bc05249b1cffba2bfb6", "local_html_sha256": "01d0b477ce4b1567e5d2684a9d591a756b5a32c26a577bc05249b1cffba2bfb6", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11351v1", "result": "retrieved", "sha256": "01d0b477ce4b1567e5d2684a9d591a756b5a32c26a577bc05249b1cffba2bfb6" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11345", "title": "SynCo: Data Synthesis Co-Training for Self-Evolving LLMs via Multi-Agent Reinforcement Learning", "published": "2026-10-08T06:40:18Z", "reviews": [ { "id": "2610.11345", "title": "SynCo: Data Synthesis Co-Training for Self-Evolving LLMs via Multi-Agent Reinforcement Learning", "published": "2026-10-08T06:40:18Z", "decision": "reviewed_deferred", "reason": "已读SynCo方法/实验：独立参数Synthesizer/Reasoner joint MARL，multirollout outcome给correctness/quality/reliability/teachability互补reward，8mathbench与FrozenSynth控制。下一批需两个真实trainable模型闭环，staticpool或固定困难题非核心实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11345v1", "sha256": "79aa7675f65237f134ede261da25ad83100d309d0faf03e4b3bf73efe37a8c31", "local_html_sha256": "79aa7675f65237f134ede261da25ad83100d309d0faf03e4b3bf73efe37a8c31", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11345v1", "result": "retrieved", "sha256": "79aa7675f65237f134ede261da25ad83100d309d0faf03e4b3bf73efe37a8c31" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11334", "title": "ReCast: Attribution-Oriented Step Representation Learning for LLM-Based Agent Systems", "published": "2026-10-08T06:31:37Z", "reviews": [ { "id": "2610.11334", "title": "ReCast: Attribution-Oriented Step Representation Learning for LLM-Based Agent Systems", "published": "2026-10-08T06:31:37Z", "decision": "reviewed_deferred", "reason": "已读ReCast方法/实验：冻结LM hidden state选择layer、pattern/deviation特征，contextencoder contrastive+ranking训练，ReCast2K与task-grouped heldout。下一批需真实activation及encoder训练，测试rootcause标签只外部分数，anonymous repo尚待审。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 METHODOLOGY", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11334v1", "sha256": "2ce680795782786b55b3cc592372b57dc63ff6ef1230a6c4bf2e5b0a02ff763a", "local_html_sha256": "2ce680795782786b55b3cc592372b57dc63ff6ef1230a6c4bf2e5b0a02ff763a", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11334v1", "result": "retrieved", "sha256": "2ce680795782786b55b3cc592372b57dc63ff6ef1230a6c4bf2e5b0a02ff763a" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11332", "title": "ReCal: Calibrating Structured Pruning for On-Policy Distillation Recovery", "published": "2026-10-08T06:28:38Z", "reviews": [ { "id": "2610.11332", "title": "ReCal: Calibrating Structured Pruning for On-Policy Distillation Recovery", "published": "2026-10-08T06:28:38Z", "decision": "accepted", "reason": "ReCal已有实现：teacher与pruned probe前向KL重权calibration，引导现有结构剪枝并SFT→OPD恢复；六benchmark需保持实剪/恢复接口，本次原文复核不声称新增运行。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "4 Recovery-Aware Calibration for Pruning", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11332v1", "sha256": "2ec4f8c652d52479e2604f4e976dccc1c5b93cf137c10ee56ce131b918e78493", "local_html_sha256": "2ec4f8c652d52479e2604f4e976dccc1c5b93cf137c10ee56ce131b918e78493", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11332v1", "result": "retrieved", "sha256": "2ec4f8c652d52479e2604f4e976dccc1c5b93cf137c10ee56ce131b918e78493" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "accepted" }, { "id": "2610.11329", "title": "PathLang: A Language-Centered Benchmark for Vision-Language Models in Computational Pathology", "published": "2026-10-08T06:27:42Z", "reviews": [ { "id": "2610.11329", "title": "PathLang: A Language-Centered Benchmark for Vision-Language Models in Computational Pathology", "published": "2026-10-08T06:27:42Z", "decision": "reviewed_deferred", "reason": "PathLang冻结slides/labels及image-textdirection，只变临床diagnosticphrase：zero-shot分类、crossmodalretrieval、semanticallyequivalent/length/style paraphrase、promptensemble、四竞争candidatepool。九patchdualencoders queryconditionedtopKtilepool统一slideinterface，并对四ratio和meanpool敏感性分析，六boardcertifiedpathologist审prompt/pool。; PANDA10609/UniToPatho237/CAMELYON16 399/CAMELYON17 499/TCGAGBMLGG1335、四organ；临床等义改写高predictioninstability，meanpool仍存在。主表每model/dataset(/pool)按evaluationlabel选最好aggregation，T3同配置固定全部variant/ensembles，所以只是oracleperformance reference。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11329v1#S3", "https://arxiv.org/html/2610.11329v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11329v1", "sha256": "aebbd58b50f92b093f0eb4a3d55baf825c3490b771d7ee0077bcc565274ac580" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11317", "title": "DivMoE: Fine-Grained MoE Upcycling via Cross-Domain Expert Composition", "published": "2026-10-08T06:23:01Z", "reviews": [ { "id": "2610.11317", "title": "DivMoE: Fine-Grained MoE Upcycling via Cross-Domain Expert Composition", "published": "2026-10-08T06:23:01Z", "decision": "reviewed_deferred", "reason": "DivMoE从同base进行math/code/science/common四domainfullsequenceCPT，再每SwiGLUFFN gate/up按row、down按column配对切两shards为八fineexperts，magnitudecorrectiongamma=(E*G²/k)^(1/3)，nonexpertweights取base+specialistsmean。router各domain选top1，再跨group选topk=2、hard atmostone/group，softmaxweight仍非equal，随后统一Stage2CPT/Stage3SFT。; Qwen3-1.7BBase/Llama3.2-1B及4B→12B，15bench六upcyclingbaseline；Stage2所有500B tokens，BTX/DivMoE另Stage1共40B、总540B。Qwen1.7平均55.6vsbestbaseline51.6，Dropfine23.2nearfromscratch22.2；publicTulu3SFT同应用matchedvariants，12B全8bench64.5=Moonlight16B。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11317v1#S3", "https://arxiv.org/html/2610.11317v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11317v1", "sha256": "6ff361abbb5a24b9fff435cc3a67a5fdd33dc062663411d5a3bc1eb3bfb3b031" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11316", "title": "MetaEncoder: Exploring the Limit of Bi-Encoders for Multimodal System One Decision Making with Natural Language Interface", "published": "2026-10-08T06:22:48Z", "reviews": [ { "id": "2610.11316", "title": "MetaEncoder: Exploring the Limit of Bi-Encoders for Multimodal System One Decision Making with Natural Language Interface", "published": "2026-10-08T06:22:48Z", "decision": "reviewed_deferred", "reason": "Muse-Glimmer30B causal last-token单位范数bi-encoder，request/candidate共享网络；closed-set可将option list放request，open-set缓存独立candidate；单向InfoNCE，重复positive不作negative。; §4 11 suites/190 tasks；官方eval repo与30B模型card可访问，retrieval与decision按不同完整candidate sets评测。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract/author information/main defining methods/main experiments and tables; production-evidence paragraph spot check; appendices not exhaustively read", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11316v1#S3", "https://arxiv.org/html/2610.11316v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11316v1", "sha256": "a838ee38f50a178939e66545558bd0c8eb5298604f13f3cd7a146207b972c608" }, "new_implementation_inferred_from_review": false }, { "id": "2610.11316", "title": "MetaEncoder: Exploring the Limit of Bi-Encoders for Multimodal System One Decision Making with Natural Language Interface", "published": "2026-10-08T06:22:48Z", "decision": "reviewed_deferred", "reason": "Meta AI自然语言SystemOne双塔决策编码器，Muse-Glimmer30B与单向contrastive候选对齐，190任务/11suite需真实checkpoint候选检索实现和GPU评测；保留下一批优先。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "4 Experiment", "location": ".cache/oct12-scan/fulltext/2610.11316.json sections text prefix", "characters_read": [ 0, 1200 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 MetaEncoder", "section_index": 2, "characters_read": [ 0, 1604 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 139, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11316v1", "sha256": "a838ee38f50a178939e66545558bd0c8eb5298604f13f3cd7a146207b972c608", "local_path": ".cache/oct12-scan/fulltext/2610.11316.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11316v1", "result": "retrieved", "sha256": "a838ee38f50a178939e66545558bd0c8eb5298604f13f3cd7a146207b972c608" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11316", "sha256": "cde8e1c298234a0431759e13d80cf4d633d7863825994c1856dac9f0ba2ac86a", "local_path": ".cache/oct12-scan/fulltext/2610.11316.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 06:22:48 UTC (324 KB)", "published": "2026-10-08T06:22:48Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11314", "title": "From Retrieval to Reconstruction: Constructing Evolvable Cognitive Memory for Long-Term Dialogue", "published": "2026-10-08T06:21:45Z", "reviews": [ { "id": "2610.11314", "title": "From Retrieval to Reconstruction: Constructing Evolvable Cognitive Memory for Long-Term Dialogue", "published": "2026-10-08T06:21:45Z", "decision": "reviewed_deferred", "reason": "PEC²F graph将主观Claim与Fact/Event分离，保留source/target和SUPPORTED_BY；同source/predicate按时间reconcile，跨speaker不合并；四deterministic operators由LLM意图控制。; LoCoMo与LongMemEval两英文backbone、semantic-collapse synthetic probe、50样本人偏好；offline consolidation成本/graph增长。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11314v1#S3", "https://arxiv.org/html/2610.11314v1#S4", "https://arxiv.org/html/2610.11314v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11314v1", "sha256": "7b5396b30f5fc13726c35857d533598014cfb35a35fee84115c185efebf8063e" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11312", "title": "MedBenchAgent: Towards Systematic Automation of Medical VLM Benchmark Construction", "published": "2026-10-08T06:17:33Z", "reviews": [ { "id": "2610.11312", "title": "MedBenchAgent: Towards Systematic Automation of Medical VLM Benchmark Construction", "published": "2026-10-08T06:17:33Z", "decision": "reviewed_deferred", "reason": "已读MedBenchAgent BIR方法/benchmark定义：五agent锁spec后instantiate，task/evidence/evaluation协议约束编译，994/1000人工audit为正确task条件下数值。下一批benchmarkinfra须医学专家审查与annotation authority，不能由生成judge自证可靠。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 What Makes a Good Medical VLM Benchmark?", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 MedBenchAgent", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11312v1", "sha256": "cb234b98bc850fd1c6a3d7afd7945908419b5c734d38a4db634a140105da874b", "local_html_sha256": "cb234b98bc850fd1c6a3d7afd7945908419b5c734d38a4db634a140105da874b", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11312v1", "result": "retrieved", "sha256": "cb234b98bc850fd1c6a3d7afd7945908419b5c734d38a4db634a140105da874b" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11310", "title": "FloorSAV: Elucidating Spatial Audio-Visual Context with 2D Floormap for AV-LLMs", "published": "2026-10-08T06:15:36Z", "reviews": [ { "id": "2610.11310", "title": "FloorSAV: Elucidating Spatial Audio-Visual Context with 2D Floormap for AV-LLMs", "published": "2026-10-08T06:15:36Z", "decision": "reviewed_deferred", "reason": "FloorSAV把点云/轨迹/声源/landmark同步2D地图注入AV-LLM；SAVED/SAVVY需区分估计与GT地图，真实multimodal数据和推理未在本批实施。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Method: FloorSAV", "location": ".cache/oct12-scan/fulltext/2610.11310.json sections text prefix", "characters_read": [ 0, 950 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Spatial Reasoning Benchmarks", "location": ".cache/oct12-scan/fulltext/2610.11310.json sections text prefix", "characters_read": [ 0, 1200 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "5 Experiments and Results", "section_index": 4, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 140, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11310v1", "sha256": "bb20ee4e8ed822df91200075b53c39a7405212665e45a18553e3a4964f5c3965", "local_path": ".cache/oct12-scan/fulltext/2610.11310.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11310v1", "result": "retrieved", "sha256": "bb20ee4e8ed822df91200075b53c39a7405212665e45a18553e3a4964f5c3965" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11310", "sha256": "2feeb4e9a7b0ebcddf24294c9df58d872b822d600bc23512f5e90a80698b31e8", "local_path": ".cache/oct12-scan/fulltext/2610.11310.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 06:15:36 UTC (12,737 KB)", "published": "2026-10-08T06:15:36Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11303", "title": "Efficient Multi-Granularity Knowledge Transfer for Radiology Report Generation", "published": "2026-10-08T06:09:28Z", "reviews": [ { "id": "2610.11303", "title": "Efficient Multi-Granularity Knowledge Transfer for Radiology Report Generation", "published": "2026-10-08T06:09:28Z", "decision": "reviewed_deferred", "reason": "EMGKT全局medical VLM+finegrained teacher/student蒸馏+疾病expert，IU-Xray/MIMIC-CXR公开评测有访问/跨域协议约束；不把临床报告NLG代理当临床能力。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "III Method", "location": ".cache/oct12-scan/fulltext/2610.11303.json sections text prefix", "characters_read": [ 0, 950 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "IV Experiment", "location": ".cache/oct12-scan/fulltext/2610.11303.json sections text prefix", "characters_read": [ 0, 1200 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 141, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11303v1", "sha256": "5d96cbf916242bb883c428be3382dc0632f678770ee878af2057b1800c83c5e6", "local_path": ".cache/oct12-scan/fulltext/2610.11303.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11303v1", "result": "retrieved", "sha256": "5d96cbf916242bb883c428be3382dc0632f678770ee878af2057b1800c83c5e6" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11303", "sha256": "7451df50ed7d076f5036bbff6e58701b32418b6e63f3684481699b76ce5816b8", "local_path": ".cache/oct12-scan/fulltext/2610.11303.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 06:09:28 UTC (990 KB)", "published": "2026-10-08T06:09:28Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11299", "title": "DuplexAgent-RSI: Recursive Harness Improvement for Full-Duplex Voice Agent Collaboration", "published": "2026-10-08T06:06:42Z", "reviews": [ { "id": "2610.11299", "title": "DuplexAgent-RSI: Recursive Harness Improvement for Full-Duplex Voice Agent Collaboration", "published": "2026-10-08T06:06:42Z", "decision": "reviewed_deferred", "reason": "已读DuplexAgent方法/实验：固定duplexmodel+async delegation，六可编辑harness模块、timed simulator failure与ExamPlanner/Editor archive闭环。external spoken/tool任务与simulator editablation分开，下一批需真实时序取消/替换/交付及音频model，不是同步textfixture。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11299v1", "sha256": "0782a5ab5f7cf689360499aa8d06e511100a3f7fda24f07a9c5760b47b98ca27", "local_html_sha256": "0782a5ab5f7cf689360499aa8d06e511100a3f7fda24f07a9c5760b47b98ca27", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11299v1", "result": "retrieved", "sha256": "0782a5ab5f7cf689360499aa8d06e511100a3f7fda24f07a9c5760b47b98ca27" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11291", "title": "When Do We Need On-Policy Distillation? Distilling on Offline Student Rollouts Is Often Better", "published": "2026-10-08T05:54:03Z", "reviews": [ { "id": "2610.11291", "title": "When Do We Need On-Policy Distillation? Distilling on Offline Student Rollouts Is Often Better", "published": "2026-10-08T05:54:03Z", "decision": "accepted", "reason": "Semi-OPD已implemented，固定initial student离线rollout teacher token cache，以teacher/student overlap诊断何时持续OPD；17 pairs对照覆盖准确率/效率，本次保留既有而非新重复实施。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "2 An Efficient OPD Alternative: Semi-On-Policy Distillation", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "2 An Efficient OPD Alternative: Semi-On-Policy Distillation", "scope": "additional Semi-OPD formula, caching and paired-model results excerpt", "characters_read_at_least": [ 1300, 6500 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11291v1", "sha256": "7d840425b74c01b525bbb24d7b11ce46ee153db415e38ea03f00ebac6e330f21", "local_html_sha256": "7d840425b74c01b525bbb24d7b11ce46ee153db415e38ea03f00ebac6e330f21", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11291v1", "result": "retrieved", "sha256": "7d840425b74c01b525bbb24d7b11ce46ee153db415e38ea03f00ebac6e330f21" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "accepted" }, { "id": "2610.11287", "title": "REMORY: Learning Residual Memory for Context Compaction", "published": "2026-10-08T05:50:50Z", "reviews": [ { "id": "2610.11287", "title": "REMORY: Learning Residual Memory for Context Compaction", "published": "2026-10-08T05:50:50Z", "decision": "reviewed_deferred", "reason": "summary-conditioned neural memory产生bounded soft tokens附summary后；frozen full-history teacher与compacted student的complete-vocab reverse-KL，stageI reconstruction初始化、stageII residual objective，仅memory phi训练。; SummHay全部92query、额外位置等预算summary++、多longhorizonagent任务；失败保留，gold insight/citations仅scoring。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11287v1#S2", "https://arxiv.org/html/2610.11287v1#S3" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11287v1", "sha256": "4ccf9c975837526fc3d5c91f84407b2baf52fe23aa5a98e90070604ecf8af5ac" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11281", "title": "How to post-train on a surrogate: Envelope sampling mitigates reward hacking", "published": "2026-10-08T05:48:31Z", "reviews": [ { "id": "2610.11281", "title": "How to post-train on a surrogate: Envelope sampling mitigates reward hacking", "published": "2026-10-08T05:48:31Z", "decision": "reviewed_deferred", "reason": "Envelope sampling在judge L2 uncertainty ball下优化regret bound，用reject或modified-reward finetune采样标注recalibration，clinical notes/sycophancy反rewardhacking；下一批需expert labels与heldout calibration边界，不能oracle答案给被评估policy。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "Theoretical results", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11281v1", "sha256": "0702714fd93108be1cc5bf2b54efe4b4df4141eef22d6a5616d3b7db3a497451", "local_html_sha256": "0702714fd93108be1cc5bf2b54efe4b4df4141eef22d6a5616d3b7db3a497451", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11281v1", "result": "retrieved", "sha256": "0702714fd93108be1cc5bf2b54efe4b4df4141eef22d6a5616d3b7db3a497451" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11253", "title": "LLM-IDEA: Identifiability-Driven Experimental Agent for Autonomous Discovery of Mechanistic World Models", "published": "2026-10-08T05:01:30Z", "reviews": [ { "id": "2610.11253", "title": "LLM-IDEA: Identifiability-Driven Experimental Agent for Autonomous Discovery of Mechanistic World Models", "published": "2026-10-08T05:01:30Z", "decision": "reviewed_deferred", "reason": "已读LLM-IDEA council/method及results：identifiability engine三类plateau verdict与合法experiment认证，benchmark可奖励不可识别解释；三角色court实际零modelcall，sharedknowledge可能破坏seed独立。下一批需真实symbolic engine与trial，不按8seed名称误认独立重复。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11253v1", "sha256": "5a7f94f33c4fbcf7253e4812686012b6e7fa91e692ac13dc88b111f89602ecb8", "local_html_sha256": "5a7f94f33c4fbcf7253e4812686012b6e7fa91e692ac13dc88b111f89602ecb8", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11253v1", "result": "retrieved", "sha256": "5a7f94f33c4fbcf7253e4812686012b6e7fa91e692ac13dc88b111f89602ecb8" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11251", "title": "V-CoLA: Vision Token Compression with Linear Attention", "published": "2026-10-08T04:55:50Z", "reviews": [ { "id": "2610.11251", "title": "V-CoLA: Vision Token Compression with Linear Attention", "published": "2026-10-08T04:55:50Z", "decision": "accepted", "reason": "V-CoLA线性attention uniqueness importance与chunk并行adaptive merging，Qwen3.5/InfiniteVL公开多VQA/A100结果；须与既有vcola适配核对真实CUDA证据，不在本审查宣称新增实现/通过。；已有本地vcola实现与Oct11A100 L1 receipt，本closure只继承，不升级4image diagnostic为论文benchmark能力比较。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Method", "location": ".cache/oct12-scan/fulltext/2610.11251.json sections text prefix", "characters_read": [ 0, 950 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.11251.json sections text prefix", "characters_read": [ 0, 1200 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 Method", "section_index": 2, "characters_read": [ 1400, 2800 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "3 Method", "location": "text[3500:8500] actual 3.2 uniqueness score and probes", "characters_read": [ 3500, 8500 ], "scope": "bounded excerpt; raw source available, cached section extraction may be capped" }, { "role": "additional targeted method/evaluation", "heading": "3 Method", "location": "text[8500:11500] actual 3.3 partition/softmax merging and Table1", "characters_read": [ 8500, 11500 ], "scope": "bounded excerpt; raw source available, cached section extraction may be capped" }, { "role": "additional targeted method/evaluation", "heading": "4 Experiments", "location": "text[1200:2900] hyperparameters/actual component runtime", "characters_read": [ 1200, 2900 ], "scope": "bounded excerpt; raw source available, cached section extraction may be capped" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 142, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11251v1", "sha256": "b96430ee5861209868a532880df8c29e4ad049b0b77858a73a920f8085ea2001", "local_path": ".cache/oct12-scan/fulltext/2610.11251.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11251v1", "result": "retrieved", "sha256": "b96430ee5861209868a532880df8c29e4ad049b0b77858a73a920f8085ea2001" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11251", "sha256": "092c90817f3da4bac3cb7cb343232372304b97befcdec292887daeb5919567f5", "local_path": ".cache/oct12-scan/fulltext/2610.11251.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 04:55:50 UTC (4,293 KB)", "published": "2026-10-08T04:55:50Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "accepted" }, { "id": "2610.11247", "title": "Why On-Policy Distillation Sometimes Fails: Vanishing Learning Signals", "published": "2026-10-08T04:52:05Z", "reviews": [ { "id": "2610.11247", "title": "Why On-Policy Distillation Sometimes Fails: Vanishing Learning Signals", "published": "2026-10-08T04:52:05Z", "decision": "accepted", "reason": "OPD-learning-signals已implemented diagnostic：small-LR dynamics与true logs gradient proxy、near-student teacher条件保证；early plateau/CKA相关性不证明原因，保持diagnostic而非capability method。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Preliminaries", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Early Loss Plateaus in On-Policy Distillation", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11247v1", "sha256": "614cf91e0d89f57fedbb7e4d6494e2c96cb2e945bdb7edb63df2399b023c2e05", "local_html_sha256": "614cf91e0d89f57fedbb7e4d6494e2c96cb2e945bdb7edb63df2399b023c2e05", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11247v1", "result": "retrieved", "sha256": "614cf91e0d89f57fedbb7e4d6494e2c96cb2e945bdb7edb63df2399b023c2e05" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "accepted" }, { "id": "2610.11245", "title": "Read What Matters: Query-Adaptive Quantization for KV Caches", "published": "2026-10-08T04:47:41Z", "reviews": [ { "id": "2610.11245", "title": "Read What Matters: Query-Adaptive Quantization for KV Caches", "published": "2026-10-08T04:47:41Z", "decision": "reviewed_deferred", "reason": "ReadKV progressive存储/每query keychannel再valuetoken读bit分配，区分retained/read预算，六base C4+longQA和A10G单layer测量；需真实partial-fetch kernel而非全量解码后mask。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "2 Partial-Read Cache Model", "location": ".cache/oct12-scan/fulltext/2610.11245.json sections text prefix", "characters_read": [ 0, 950 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Performance Guarantees for Partial Reads", "location": ".cache/oct12-scan/fulltext/2610.11245.json sections text prefix", "characters_read": [ 0, 1200 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 Query-Adaptive Cache Design", "section_index": 2, "characters_read": [ 0, 1406 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "5 Language-Model Evaluation", "section_index": 4, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 143, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11245v1", "sha256": "ce870e76de345881f2c704562845cd02c7031fcadc9b49e47b59491367a7dc81", "local_path": ".cache/oct12-scan/fulltext/2610.11245.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11245v1", "result": "retrieved", "sha256": "ce870e76de345881f2c704562845cd02c7031fcadc9b49e47b59491367a7dc81" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11245", "sha256": "6bf44476733691887a6b6e7c658117c5b3b1c28fa0c9a050a711ad6519715962", "local_path": ".cache/oct12-scan/fulltext/2610.11245.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 04:47:41 UTC (117 KB)", "published": "2026-10-08T04:47:41Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11231", "title": "Harness Compilation: Which Decisions Should a Small Vision-Language Model Keep?", "published": "2026-10-08T04:31:36Z", "reviews": [ { "id": "2610.11231", "title": "Harness Compilation: Which Decisions Should a Small Vision-Language Model Keep?", "published": "2026-10-08T04:31:36Z", "decision": "reviewed_deferred", "reason": "Harness Compilation冻结student，从execution traces由teacher编译内容/控制，再独立validation选部署harness；7视觉任务3build且outergate预算明确，下一批需可执行hook和隔离反馈。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "4 Harness Compilation", "location": ".cache/oct12-scan/fulltext/2610.11231.json sections text prefix", "characters_read": [ 0, 950 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5 Experiments", "location": ".cache/oct12-scan/fulltext/2610.11231.json sections text prefix", "characters_read": [ 0, 1200 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 144, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11231v1", "sha256": "dd1eca1d377f540a5794feb0ab8c9efafac9479579a700f469dca3f092a46079", "local_path": ".cache/oct12-scan/fulltext/2610.11231.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11231v1", "result": "retrieved", "sha256": "dd1eca1d377f540a5794feb0ab8c9efafac9479579a700f469dca3f092a46079" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11231", "sha256": "f1f1461e6426ebe9f953b252fb82825fd9812c934e1fd275db5fa68bfa39b4a2", "local_path": ".cache/oct12-scan/fulltext/2610.11231.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 04:31:36 UTC (379 KB)", "published": "2026-10-08T04:31:36Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11226", "title": "When Lower Reconstruction Loss Hurts: Distributionally Robust Refinement for Low-Bit LLM Quantization", "published": "2026-10-08T04:22:29Z", "reviews": [ { "id": "2610.11226", "title": "When Lower Reconstruction Loss Hurts: Distributionally Robust Refinement for Low-Bit LLM Quantization", "published": "2026-10-08T04:22:29Z", "decision": "reviewed_deferred", "reason": "DRQ固定quant grid只refine integer codes，worst-case activation distributions纠正calibration reconstruction泛化，六PTQ methods dense/MoE公开ppl/tasks；下一批须真实整数码/量化forward/uncertainty inner problem。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 DRQ: Distributionally Robust Quantization", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experimental Evaluation", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11226v1", "sha256": "e4a077366e7ad0dd0c0dc82e9a3e1d78dc26353c7e10ef4ae50750ec58a8ef27", "local_html_sha256": "e4a077366e7ad0dd0c0dc82e9a3e1d78dc26353c7e10ef4ae50750ec58a8ef27", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11226v1", "result": "retrieved", "sha256": "e4a077366e7ad0dd0c0dc82e9a3e1d78dc26353c7e10ef4ae50750ec58a8ef27" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11216", "title": "The Lattice of Transition Laws", "published": "2026-10-08T04:13:05Z", "reviews": [ { "id": "2610.11216", "title": "The Lattice of Transition Laws", "published": "2026-10-08T04:13:05Z", "decision": "reviewed_deferred", "reason": "corruption lattice把AR/diffusion schedule依赖丢失成本联系treedepth，matchedstep text8/MAR/LLaDA/video验证排序；理论+采样调度真实模型实现留下一批。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "2 Decoding Schedules on the Corruption Lattice", "location": ".cache/oct12-scan/fulltext/2610.11216.json sections text prefix", "characters_read": [ 0, 950 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "6 Experiments", "location": ".cache/oct12-scan/fulltext/2610.11216.json sections text prefix", "characters_read": [ 0, 1200 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 145, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11216v1", "sha256": "9c0bc83574e129f9cfbb5cbd0bee2dc38a35c5cc16fe89c5d65219f2e8281662", "local_path": ".cache/oct12-scan/fulltext/2610.11216.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11216v1", "result": "retrieved", "sha256": "9c0bc83574e129f9cfbb5cbd0bee2dc38a35c5cc16fe89c5d65219f2e8281662" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11216", "sha256": "235542501a9c89282637170b1521547416e02b99fe40655625998ab31696d6d7", "local_path": ".cache/oct12-scan/fulltext/2610.11216.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 04:13:05 UTC (408 KB)", "published": "2026-10-08T04:13:05Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11214", "title": "Bridging KV-Cache Quantization and Linear Attention: From Theory to Pretrained Weight Migration", "published": "2026-10-08T04:12:49Z", "reviews": [ { "id": "2610.11214", "title": "Bridging KV-Cache Quantization and Linear Attention: From Theory to Pretrained Weight Migration", "published": "2026-10-08T04:12:49Z", "decision": "reviewed_deferred", "reason": "RAM-Net软离散地址和recurrent slot桥接quantized matching/linear aggregation，并500M FineWebEdu迁移九teacher；需完整多stage真实权重迁移与GPU评测。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Bridging Discrete-Code Soft-NN and Continuous Linear Predictors", "location": ".cache/oct12-scan/fulltext/2610.11214.json sections text prefix", "characters_read": [ 0, 950 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5 Experiments", "location": ".cache/oct12-scan/fulltext/2610.11214.json sections text prefix", "characters_read": [ 0, 1200 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 146, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11214v1", "sha256": "7f9c62059d83f68ebd6a412e83c3369dce28c743be8c01832df90b8b921198ff", "local_path": ".cache/oct12-scan/fulltext/2610.11214.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11214v1", "result": "retrieved", "sha256": "7f9c62059d83f68ebd6a412e83c3369dce28c743be8c01832df90b8b921198ff" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11214", "sha256": "461251d45825207a1b9e10a2a4f5db863cee37e006a454bd01991958be1187ff", "local_path": ".cache/oct12-scan/fulltext/2610.11214.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 04:12:49 UTC (10,602 KB)", "published": "2026-10-08T04:12:49Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11205", "title": "Dissecting Representation Structure in Vision Transformers: A Rigorous Architectural Study", "published": "2026-10-08T04:06:50Z", "reviews": [ { "id": "2610.11205", "title": "Dissecting Representation Structure in Vision Transformers: A Rigorous Architectural Study", "published": "2026-10-08T04:06:50Z", "decision": "reviewed_deferred", "reason": "ViT module feature entropy/min eigenvalue代理与1000 AutoFormer小模型ImageNet generalization ranking；保留architecture分析候选，NAS应validation选择避免拿test排名当选择收益。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "V Discovering the Impact of Feature Information on ViT Generalization Predictors", "location": ".cache/oct12-scan/fulltext/2610.11205.json sections text prefix", "characters_read": [ 0, 950 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "VI Efficient ViT Architecture Design", "section_index": 5, "characters_read": [ 0, 1356 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 147, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11205v1", "sha256": "1f0094c9e5d41854bc86d3ecd7b3850cdab11983b11056b0caa208b52ce63223", "local_path": ".cache/oct12-scan/fulltext/2610.11205.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11205v1", "result": "retrieved", "sha256": "1f0094c9e5d41854bc86d3ecd7b3850cdab11983b11056b0caa208b52ce63223" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11205", "sha256": "7fa74dceddc783a6ca9ee36af6959cff47e5dec93ad5c4b9d4061470c3b5d459", "local_path": ".cache/oct12-scan/fulltext/2610.11205.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 04:06:50 UTC (531 KB)", "published": "2026-10-08T04:06:50Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11196", "title": "Selective Listening: Mechanism-Guided Control of Audio Influence in Large Audio-Language Models", "published": "2026-10-08T04:00:47Z", "reviews": [ { "id": "2610.11196", "title": "Selective Listening: Mechanism-Guided Control of Audio Influence in Large Audio-Language Models", "published": "2026-10-08T04:00:47Z", "decision": "reviewed_deferred", "reason": "ICAP-Gate模型特定lateaudio路径按audio-demand task控制；四LALM ARC/MMLU×两干扰与ASR保真、pairedflip/latency；下一批需模型真实路径干预和配对gold隔离。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Methodology", "location": ".cache/oct12-scan/fulltext/2610.11196.json sections text prefix", "characters_read": [ 0, 950 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.11196.json sections text prefix", "characters_read": [ 0, 1200 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 148, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11196v1", "sha256": "f20d476d96ee2185468bd9034fbc43b39aa1e0fe38c8ec46dfc074e797f4fff9", "local_path": ".cache/oct12-scan/fulltext/2610.11196.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11196v1", "result": "retrieved", "sha256": "f20d476d96ee2185468bd9034fbc43b39aa1e0fe38c8ec46dfc074e797f4fff9" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11196", "sha256": "e2ae92de5a276fd5943bd739b527e3338e99b0b269a46071a526719a271f002c", "local_path": ".cache/oct12-scan/fulltext/2610.11196.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 04:00:47 UTC (2,253 KB)", "published": "2026-10-08T04:00:47Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11188", "title": "What to Admit and How to Present: Governing Persistent Memory in LLM Agents", "published": "2026-10-08T03:45:28Z", "reviews": [ { "id": "2610.11188", "title": "What to Admit and How to Present: Governing Persistent Memory in LLM Agents", "published": "2026-10-08T03:45:28Z", "decision": "reviewed_deferred", "reason": "已读memory admission/presentation方法/实验：FC整entry与PS typedunits、确定性permission决策，matchedbudget与Holm校正；降低leak却增personalization failures，预注册criteria未全满足。下一批保留负结果，不宣称整体personalization有效或随意删memory。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4. Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3. Separating Admission from Presentation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11188v1", "sha256": "263cc715331c0f3824bf493c0f88e562cf9c7e7937211be08b9c9f936dddc8fa", "local_html_sha256": "263cc715331c0f3824bf493c0f88e562cf9c7e7937211be08b9c9f936dddc8fa", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11188v1", "result": "retrieved", "sha256": "263cc715331c0f3824bf493c0f88e562cf9c7e7937211be08b9c9f936dddc8fa" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11183", "title": "RAG-Stress: Probing the Limits of Evidence Reliance in Retrieval-Augmented Generation", "published": "2026-10-08T03:40:48Z", "reviews": [ { "id": "2610.11183", "title": "RAG-Stress: Probing the Limits of Evidence Reliance in Retrieval-Augmented Generation", "published": "2026-10-08T03:40:48Z", "decision": "reviewed_deferred", "reason": "RAG-Stress对同question/reference的单断言干预，baseline-correct subset测misleading rate；这是证据依赖诊断，下一批转evaluation，不作后训练改进算法或能力实施证据。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 RAG-Stress Methodology", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11183v1", "sha256": "c707584dc1f93b78923e66dbdb5e8e69e0b01159473db66447a2b581ca902cb2", "local_html_sha256": "c707584dc1f93b78923e66dbdb5e8e69e0b01159473db66447a2b581ca902cb2", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11183v1", "result": "retrieved", "sha256": "c707584dc1f93b78923e66dbdb5e8e69e0b01159473db66447a2b581ca902cb2" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11179", "title": "Who Pays the Review Cost? Triage, Fairness, and Accountability in AI-authored Pull Requests", "published": "2026-10-08T03:38:29Z", "reviews": [ { "id": "2610.11179", "title": "Who Pays the Review Cost? Triage, Fairness, and Accountability in AI-authored Pull Requests", "published": "2026-10-08T03:38:29Z", "decision": "rejected_or_out_of_scope", "reason": "已读review-cost survey方法/结果：239 practitioner31国混合问卷，提出reviewability debt与ownership/rationale规范，不是可执行新agent算法/bench。真实人的偏好调查可作工程参考，不当模型因果效果。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Evaluation and Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11179v1", "sha256": "fbac7a42c1a275b4c86c43a00589f366d8660d7e1c20be70c52ca6bbaa334332", "local_html_sha256": "fbac7a42c1a275b4c86c43a00589f366d8660d7e1c20be70c52ca6bbaa334332", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11179v1", "result": "retrieved", "sha256": "fbac7a42c1a275b4c86c43a00589f366d8660d7e1c20be70c52ca6bbaa334332" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.11169", "title": "Skill Constellations: Tracing the Supply Chain of Agent Skills on GitHub", "published": "2026-10-08T03:25:22Z", "reviews": [ { "id": "2610.11169", "title": "Skill Constellations: Tracing the Supply Chain of Agent Skills on GitHub", "published": "2026-10-08T03:25:22Z", "decision": "reviewed_deferred", "reason": "已读Skill Constellations数据/结果：git历史datedcopy network和未来adoption预测audit ranking，2.19m skill copies；sourcefix很少同步，stars不是源影响。下一批供应链评测infra应版本/来源固定，预测减少高riskadoption不是已部署阻断效果。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "II Data and Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "III Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11169v1", "sha256": "efe749266107dad56d9b2e628f9a01ff0862982158d1fdebc1eca7f05eac21e3", "local_html_sha256": "efe749266107dad56d9b2e628f9a01ff0862982158d1fdebc1eca7f05eac21e3", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11169v1", "result": "retrieved", "sha256": "efe749266107dad56d9b2e628f9a01ff0862982158d1fdebc1eca7f05eac21e3" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11167", "title": "PIVOT: Perplexity-Informed KD-to-RL Transition Scheduling for Vertical-Domain Few-Shot Distillation", "published": "2026-10-08T03:24:55Z", "reviews": [ { "id": "2610.11167", "title": "PIVOT: Perplexity-Informed KD-to-RL Transition Scheduling for Vertical-Domain Few-Shot Distillation", "published": "2026-10-08T03:24:55Z", "decision": "accepted", "reason": "PIVOT已有实现：teacher sequence perplexity per-sample OPD→GRPO切换，在Banking77/HWU64 fewshot同postwarmup student steps比较；原文明确train seed与officialtest，保留existing，追加实验不混选test。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "2 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11167v1", "sha256": "16faf341f9b98e02844d8185693c43fe02019cc26f828f2b698ebfffeed54399", "local_html_sha256": "16faf341f9b98e02844d8185693c43fe02019cc26f828f2b698ebfffeed54399", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11167v1", "result": "retrieved", "sha256": "16faf341f9b98e02844d8185693c43fe02019cc26f828f2b698ebfffeed54399" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "accepted" }, { "id": "2610.11161", "title": "VGGTWorld-VLA: Intent-Conditioned 3D World Evolution for Autonomous Driving", "published": "2026-10-08T03:20:48Z", "reviews": [ { "id": "2610.11161", "title": "VGGTWorld-VLA: Intent-Conditioned 3D World Evolution for Autonomous Driving", "published": "2026-10-08T03:20:48Z", "decision": "reviewed_deferred", "reason": "VGGT未来几何tokens由历史geometry、VLA语义和给定egoaction条件化；NAVSIM/OpenScene用futureVGGT伪GT，真实worldmodel训练/规划隔离留下一批。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Method", "location": ".cache/oct12-scan/fulltext/2610.11161.json sections text prefix", "characters_read": [ 0, 1000 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.11161.json sections text prefix", "characters_read": [ 0, 1150 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 149, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11161v1", "sha256": "20bf1e92eae154516e62c46b2368bbc66c38917daddc384df1252cc415238df1", "local_path": ".cache/oct12-scan/fulltext/2610.11161.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11161v1", "result": "retrieved", "sha256": "20bf1e92eae154516e62c46b2368bbc66c38917daddc384df1252cc415238df1" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11161", "sha256": "e4f253f3c7ce04e382594f5f9fa9327c40d6b182b539aee1d5cc5ad14532cfe2", "local_path": ".cache/oct12-scan/fulltext/2610.11161.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 03:20:48 UTC (6,817 KB)", "published": "2026-10-08T03:20:48Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11152", "title": "Do LLMs Learn from Rewards in Context? : Rethinking the role of reward in In-Context Reinforcement Learning", "published": "2026-10-08T03:15:00Z", "reviews": [ { "id": "2610.11152", "title": "Do LLMs Learn from Rewards in Context? : Rethinking the role of reward in In-Context Reinforcement Learning", "published": "2026-10-08T03:15:00Z", "decision": "reviewed_deferred", "reason": "direct ICRL通过reward flip/random/remove和trajectory腐化作对照，ScienceWorld/MBPP+/HumanEval+显示提升不由reward语义驱动；适合agent memory评测下一批，固定模型推理时学习不能称参数RL复现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Preliminaries", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Reward Semantics Do Not Drive Improvement", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11152v1", "sha256": "9db3a6e41dd61d84d96ef53760602b4673b9318057a9eeeff20b7263cbba3bf5", "local_html_sha256": "9db3a6e41dd61d84d96ef53760602b4673b9318057a9eeeff20b7263cbba3bf5", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11152v1", "result": "retrieved", "sha256": "9db3a6e41dd61d84d96ef53760602b4673b9318057a9eeeff20b7263cbba3bf5" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11148", "title": "SP-DocReader: Difference-Aware Self-Play for Precise Document OCR", "published": "2026-10-08T03:12:12Z", "reviews": [ { "id": "2610.11148", "title": "SP-DocReader: Difference-Aware Self-Play for Precise Document OCR", "published": "2026-10-08T03:12:12Z", "decision": "reviewed_deferred", "reason": "LCS差异mask与完整prefix相对loss+FFL，只训OCR模块且对手刷新；Vary600k/DocVQA跨域含回退，真实selfplay/GPU未复现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Methodology", "location": ".cache/oct12-scan/fulltext/2610.11148.json sections text prefix", "characters_read": [ 0, 1000 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.11148.json sections text prefix", "characters_read": [ 0, 1150 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 150, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11148v1", "sha256": "21eea9c2883f7daa9fef4f1036b6723cc2b23ce48fdf76eb8a58b6ade53d9f85", "local_path": ".cache/oct12-scan/fulltext/2610.11148.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11148v1", "result": "retrieved", "sha256": "21eea9c2883f7daa9fef4f1036b6723cc2b23ce48fdf76eb8a58b6ade53d9f85" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11148", "sha256": "6e46a392e7bc900290d6c6351d615ac303606f1c1cb2349112d04e97d9a8167d", "local_path": ".cache/oct12-scan/fulltext/2610.11148.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 03:12:12 UTC (1,030 KB)", "published": "2026-10-08T03:12:12Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11144", "title": "Improving Image-Based Nutrition Estimation Through Multimodal Food-Item Verification and Recovery", "published": "2026-10-08T03:09:29Z", "reviews": [ { "id": "2610.11144", "title": "Improving Image-Based Nutrition Estimation Through Multimodal Food-Item Verification and Recovery", "published": "2026-10-08T03:09:29Z", "decision": "reviewed_deferred", "reason": "foodinventory/SAMproposal独立identity和geometry审核，单次recovery后blind重审；N5k分母不是itemGT，commonvalid样本/营养评测边界须保留。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Methodology", "location": ".cache/oct12-scan/fulltext/2610.11144.json sections text prefix", "characters_read": [ 0, 1000 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experimental Setup", "location": ".cache/oct12-scan/fulltext/2610.11144.json sections text prefix", "characters_read": [ 0, 1150 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "5 Results and Discussion", "section_index": 4, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 151, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11144v1", "sha256": "2afc9375cca303c73767402b1a1bbe0654565ecc9e9cce230268ffc29dee52bf", "local_path": ".cache/oct12-scan/fulltext/2610.11144.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11144v1", "result": "retrieved", "sha256": "2afc9375cca303c73767402b1a1bbe0654565ecc9e9cce230268ffc29dee52bf" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11144", "sha256": "6c91cd0912445628ecfbda671c447b6e81446c2676e6a7000342cbdaccf984cd", "local_path": ".cache/oct12-scan/fulltext/2610.11144.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 03:09:29 UTC (409 KB)", "published": "2026-10-08T03:09:29Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11140", "title": "ActiveMedAgent: Cost-Aware Trajectory Learning for Multimodal Medical Diagnosis", "published": "2026-10-08T03:06:11Z", "reviews": [ { "id": "2610.11140", "title": "ActiveMedAgent: Cost-Aware Trajectory Learning for Multimodal Medical Diagnosis", "published": "2026-10-08T03:06:11Z", "decision": "reviewed_deferred", "reason": "frozenAPI VLM诊断belief轨迹utility-cost训MLP acquisition/stop；三medicalbench60/40split，需真实证据通道/成本与临床不外推。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Method", "location": ".cache/oct12-scan/fulltext/2610.11140.json sections text prefix", "characters_read": [ 0, 1000 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.11140.json sections text prefix", "characters_read": [ 0, 1150 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 Method", "section_index": 2, "characters_read": [ 1000, 2300 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 152, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11140v1", "sha256": "e0413668f4576ddf84fb6b4b4f0e1f0eede1c847b670ec586a155a4eb799f8b0", "local_path": ".cache/oct12-scan/fulltext/2610.11140.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11140v1", "result": "retrieved", "sha256": "e0413668f4576ddf84fb6b4b4f0e1f0eede1c847b670ec586a155a4eb799f8b0" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11140", "sha256": "037b662b00c6a98fe7a180d2fe9a87e013ee06d7f685b7e89a4034a41d566a2c", "local_path": ".cache/oct12-scan/fulltext/2610.11140.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 03:06:11 UTC (459 KB)", "published": "2026-10-08T03:06:11Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11136", "title": "The \"10th Juror\": Open-Set Standpoint Screening for Bureaucratic Bias Detection", "published": "2026-10-08T03:05:08Z", "reviews": [ { "id": "2610.11136", "title": "The \"10th Juror\": Open-Set Standpoint Screening for Bureaucratic Bias Detection", "published": "2026-10-08T03:05:08Z", "decision": "reviewed_deferred", "reason": "已读MARS-Gov方法/实验：legal retrieval、open-set scout动态juror、conservative routing和rewrite语义verification，DGDB/LOCO评测。下一批须独立evidence与重写含义audit，法律背景检索不可视为法律建议/无偏性保证。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11136v1", "sha256": "851c7febe1a7207ee6e5fbf21bb12cf86fbbb2668ed1282244cce8329a405719", "local_html_sha256": "851c7febe1a7207ee6e5fbf21bb12cf86fbbb2668ed1282244cce8329a405719", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11136v1", "result": "retrieved", "sha256": "851c7febe1a7207ee6e5fbf21bb12cf86fbbb2668ed1282244cce8329a405719" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11123", "title": "When Interfaces Speak: Data-Aware Generative UI Harness for Active Interaction", "published": "2026-10-08T02:49:35Z", "reviews": [ { "id": "2610.11123", "title": "When Interfaces Speak: Data-Aware Generative UI Harness for Active Interaction", "published": "2026-10-08T02:49:35Z", "decision": "reviewed_deferred", "reason": "已读GenUI task定义/实验：ToolAgent与GUICoder、DynamicUX实际sandbox交互reward、RewardAuditor meta-rubric，UI-TAU真实publicdatabase模拟任务与4B训练。下一批须真实UI/code执行和用户输入，judge reward hacking校验；rounds问卷不当生产A/B。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "6 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 Task Definition", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11123v1", "sha256": "01ca43eae751ecca712c0e0323df287e843c6d29d29bb5ac1d8aa8cb5f766241", "local_html_sha256": "01ca43eae751ecca712c0e0323df287e843c6d29d29bb5ac1d8aa8cb5f766241", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11123v1", "result": "retrieved", "sha256": "01ca43eae751ecca712c0e0323df287e843c6d29d29bb5ac1d8aa8cb5f766241" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11113", "title": "DiscoVL: Unveiling Disentangled C ross-Modal Representation Learning via Orthogonal Adversarial Regularization for V ision-Language Models", "published": "2026-10-08T02:35:56Z", "reviews": [ { "id": "2610.11113", "title": "DiscoVL: Unveiling Disentangled C ross-Modal Representation Learning via Orthogonal Adversarial Regularization for V ision-Language Models", "published": "2026-10-08T02:35:56Z", "decision": "reviewed_deferred", "reason": "DiscoVL多branch低秩crossmodal residual与orthogonaladversarialtriplet，15分类/novel/domain评测；真实CLIP适配未实施。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Methodology", "location": ".cache/oct12-scan/fulltext/2610.11113.json sections text prefix", "characters_read": [ 0, 1000 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.11113.json sections text prefix", "characters_read": [ 0, 1150 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 153, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11113v1", "sha256": "aed397484915e3a9df519abffbdf6ab1c8f7ebcd8c8c8366535e207896d0e048", "local_path": ".cache/oct12-scan/fulltext/2610.11113.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11113v1", "result": "retrieved", "sha256": "aed397484915e3a9df519abffbdf6ab1c8f7ebcd8c8c8366535e207896d0e048" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11113", "sha256": "c2731f116de451383451c079caa1a18ab2a71b6b57e6eaf98a5aed1f5e0a0ed9", "local_path": ".cache/oct12-scan/fulltext/2610.11113.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 02:35:56 UTC (5,394 KB)", "published": "2026-10-08T02:35:56Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11111", "title": "Lapras: Latent Reasoning for Time Series Language Models", "published": "2026-10-08T02:33:37Z", "reviews": [ { "id": "2610.11111", "title": "Lapras: Latent Reasoning for Time Series Language Models", "published": "2026-10-08T02:33:37Z", "decision": "reviewed_deferred", "reason": "Lapras真实continuous latent thoughts、CoT teacher selfdistill在answer hidden states对齐；4 TSLM/5 time-series QA公开评测，trainingtrace可引用annotated labels但test learner必须仅signals/questions，下一批需真实time series encoder与latent recurrence。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiment", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11111v1", "sha256": "7cc934642f9dfd128458969a4bca9380b2c8d394c6680968d399fc351efc2e9a", "local_html_sha256": "7cc934642f9dfd128458969a4bca9380b2c8d394c6680968d399fc351efc2e9a", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11111v1", "result": "retrieved", "sha256": "7cc934642f9dfd128458969a4bca9380b2c8d394c6680968d399fc351efc2e9a" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11067", "title": "Diffusion Meta-Prompting and Steering for Generalizable Foundation Model Adaptation", "published": "2026-10-08T01:26:50Z", "reviews": [ { "id": "2610.11067", "title": "Diffusion Meta-Prompting and Steering for Generalizable Foundation Model Adaptation", "published": "2026-10-08T01:26:50Z", "decision": "reviewed_deferred", "reason": "DMP对promptrepository训练条件diffusion，trainselectedanchor steering不读testclasses；公开分类/SDXL实验，但真实prompt生成checkpoint/GPU待下一批。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.11067.json sections text prefix", "characters_read": [ 0, 1150 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 Diffusion Meta-Prompting", "section_index": 2, "characters_read": [ 0, 1601 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 154, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11067v1", "sha256": "d4cda68ecc9d87d338c6d545e71b09d6fb2583835c1b30ce8973060c1099eded", "local_path": ".cache/oct12-scan/fulltext/2610.11067.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11067v1", "result": "retrieved", "sha256": "d4cda68ecc9d87d338c6d545e71b09d6fb2583835c1b30ce8973060c1099eded" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11067", "sha256": "01ebe5954938e22c528d0a4f42ca89bc734d92ff91078feb18437c16745bb3fd", "local_path": ".cache/oct12-scan/fulltext/2610.11067.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 01:26:50 UTC (5,078 KB)", "published": "2026-10-08T01:26:50Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11064", "title": "Measuring and Mitigating Solution Mode Collapse in RLVR", "published": "2026-10-08T01:25:16Z", "reviews": [ { "id": "2610.11064", "title": "Measuring and Mitigating Solution Mode Collapse in RLVR", "published": "2026-10-08T01:25:16Z", "decision": "reviewed_deferred", "reason": "ModeBench correctness+mode verifier揭示RLVRmode collapse，ReMax每discovered correctmode存oneexemplar均匀replay；五seed三scale两RLobjectives同sampling预算但增加replayupdate需记录，下一批真实mode verifier和训练。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Re:Max: Learn from New Samples and Remember Successes", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experimental Design", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11064v1", "sha256": "c4f22dd0282c6ad09e040fc5e6a3bc17cd17a67c7024517299f03742e659062f", "local_html_sha256": "c4f22dd0282c6ad09e040fc5e6a3bc17cd17a67c7024517299f03742e659062f", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11064v1", "result": "retrieved", "sha256": "c4f22dd0282c6ad09e040fc5e6a3bc17cd17a67c7024517299f03742e659062f" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.11039", "title": "Rendering-Free Lookahead for Question-Guided Active Vision", "published": "2026-10-08T00:45:46Z", "reviews": [ { "id": "2610.11039", "title": "Rendering-Free Lookahead for Question-Guided Active Vision", "published": "2026-10-08T00:45:46Z", "decision": "reviewed_deferred", "reason": "RFL从训练期privileged3DGS未来viewanswerability两stage蒸馏；E3VS scene-disjoint377test与机器人，部署不renderfuture须保持。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "IV Method", "location": ".cache/oct12-scan/fulltext/2610.11039.json sections text prefix", "characters_read": [ 0, 1000 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "V Experiments", "location": ".cache/oct12-scan/fulltext/2610.11039.json sections text prefix", "characters_read": [ 0, 1150 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 155, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11039v1", "sha256": "e0b48f34177d51fa1fab3c53834364ff444e37dd62ca1fa13a1b8ec737db335d", "local_path": ".cache/oct12-scan/fulltext/2610.11039.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11039v1", "result": "retrieved", "sha256": "e0b48f34177d51fa1fab3c53834364ff444e37dd62ca1fa13a1b8ec737db335d" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11039", "sha256": "afcf5d6ff454c3346c99968d4aa8e315411c9fcc8f05e2d10a8a216026102f92", "local_path": ".cache/oct12-scan/fulltext/2610.11039.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 00:45:46 UTC (2,370 KB)", "published": "2026-10-08T00:45:46Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11033", "title": "FedAlphaEdit: Null-Space-Aligned Merging for Collaborative Knowledge Editing", "published": "2026-10-08T00:25:25Z", "reviews": [ { "id": "2610.11033", "title": "FedAlphaEdit: Null-Space-Aligned Merging for Collaborative Knowledge Editing", "published": "2026-10-08T00:25:25Z", "decision": "reviewed_deferred", "reason": "FedAlphaEdit projectedclientstats的nullspacealignservermerge；GPT2XL/Qwen编辑/保留，one-shot理想定理不等同multi-layer实际保证。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Method", "location": ".cache/oct12-scan/fulltext/2610.11033.json sections text prefix", "characters_read": [ 0, 1000 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.11033.json sections text prefix", "characters_read": [ 0, 1150 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 156, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11033v1", "sha256": "33f2aa60631d96ade7ccc46c94faae9f72d3ca15a3f26cf5e8ea9c2316c22b30", "local_path": ".cache/oct12-scan/fulltext/2610.11033.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11033v1", "result": "retrieved", "sha256": "33f2aa60631d96ade7ccc46c94faae9f72d3ca15a3f26cf5e8ea9c2316c22b30" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11033", "sha256": "745b94f97d8e26ee1805c8a17ba0639dc228061b78896c16b876b1854d75689a", "local_path": ".cache/oct12-scan/fulltext/2610.11033.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 00:25:25 UTC (54 KB)", "published": "2026-10-08T00:25:25Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11030", "title": "NOMOS: Compiling Written Policies into Statically Verified Tool-Call Gates for LLM Agents", "published": "2026-10-08T00:23:22Z", "reviews": [ { "id": "2610.11030", "title": "NOMOS: Compiling Written Policies into Statically Verified Tool-Call Gates for LLM Agents", "published": "2026-10-08T00:23:22Z", "decision": "reviewed_deferred", "reason": "已读NOMOS方法/实验：四pass一次LLMextract后静态tool-schema repair/reject，runtime facts仅来自tool/user不信agent自述，tau2/AgentDojo固定harness。下一批需真实compiler/gate与benignutility成本，结构保证不等价NLpolicy语义无遗漏。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11030v1", "sha256": "a684778e181b7969ed9d3ae68ade09b26a344bf0e61e56aad15bd5f0537d5c80", "local_html_sha256": "a684778e181b7969ed9d3ae68ade09b26a344bf0e61e56aad15bd5f0537d5c80", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11030v1", "result": "retrieved", "sha256": "a684778e181b7969ed9d3ae68ade09b26a344bf0e61e56aad15bd5f0537d5c80" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11019", "title": "Mid-Training Language Models on Raw Video", "published": "2026-10-08T00:03:39Z", "reviews": [ { "id": "2610.11019", "title": "Mid-Training Language Models on Raw Video", "published": "2026-10-08T00:03:39Z", "decision": "reviewed_deferred", "reason": "真实raw video frames经SigLIP/projector按grid+newline+time序列输入Qwen3；two-layer head以cosine distance预测next visual token，target stop-gradient；midtraining后两arm同一image-text SFT。; §4 6.99M clips/64.4B visual tokens，128H100；4video/10image/14text，0.3/0.5/1epoch轨迹与caption-target对照。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract/author information/main defining methods/main experiments and tables; production-evidence paragraph spot check; appendices not exhaustively read", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.11019v1#S3", "https://arxiv.org/html/2610.11019v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11019v1", "sha256": "3e54ff157d3b394cea7b414fb23c3a6c6dbd939c3c7bdfb55d3e6c03d8031107" }, "new_implementation_inferred_from_review": false }, { "id": "2610.11019", "title": "Mid-Training Language Models on Raw Video", "published": "2026-10-08T00:03:39Z", "decision": "reviewed_deferred", "reason": "Meta rawvideo连续nexttoken回归且targetstopgrad，再matchedimage-textSFT；64.4B视觉tokens/128H100，语言保真只对matchedarm，未实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Next-Token Prediction on Visual Tokens", "location": ".cache/oct12-scan/fulltext/2610.11019.json sections text prefix", "characters_read": [ 0, 1000 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.11019.json sections text prefix", "characters_read": [ 0, 1150 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 157, "primary_fulltext": { "url": "https://arxiv.org/html/2610.11019v1", "sha256": "3e54ff157d3b394cea7b414fb23c3a6c6dbd939c3c7bdfb55d3e6c03d8031107", "local_path": ".cache/oct12-scan/fulltext/2610.11019.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11019v1", "result": "retrieved", "sha256": "3e54ff157d3b394cea7b414fb23c3a6c6dbd939c3c7bdfb55d3e6c03d8031107" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.11019", "sha256": "40c6ac20876ca3236cab073c8cee2dc51bb2ef5881b06606515e69ab00e198c2", "local_path": ".cache/oct12-scan/fulltext/2610.11019.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Thu, 8 Oct 2026 00:03:39 UTC (360 KB)", "published": "2026-10-08T00:03:39Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.11015", "title": "Prompts versus Rules: Auditing and Controlling Speech Naturalness Behaviors in Voice User Simulators", "published": "2026-10-07T23:55:27Z", "reviews": [ { "id": "2610.11015", "title": "Prompts versus Rules: Auditing and Controlling Speech Naturalness Behaviors in Voice User Simulators", "published": "2026-10-07T23:55:27Z", "decision": "reviewed_deferred", "reason": "已读voice naturalness方法/设置：full-duplex黑盒audio模拟，prompt与确定性injection实现的disfluency/interrupt/backchannel审计，858replayed positions。下一批需测实际行为而非config声称、audio真实时序且无未经授权电话外联。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Experimental setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11015v1", "sha256": "5d650a2f4c4bcdf85e411f6abdc898976b5f2865a20f5bf4dd26e3d7e2dbe933", "local_html_sha256": "5d650a2f4c4bcdf85e411f6abdc898976b5f2865a20f5bf4dd26e3d7e2dbe933", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11015v1", "result": "retrieved", "sha256": "5d650a2f4c4bcdf85e411f6abdc898976b5f2865a20f5bf4dd26e3d7e2dbe933" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11008", "title": "Why LLM Agents Favor Their Group: Stakes, Observed Norms, and Reputation", "published": "2026-10-07T23:46:37Z", "reviews": [ { "id": "2610.11008", "title": "Why LLM Agents Favor Their Group: Stakes, Observed Norms, and Reputation", "published": "2026-10-07T23:46:37Z", "decision": "reviewed_deferred", "reason": "已读群体favoritism实验/结果：任意group、costless/stake、scripted history与个体reputation控制，4400模拟societies。属于agent社会行为测量，保留下一批参考；不能把合成群体pointallocation解释为现实组织因果。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Experimental Design", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Stakes Reveal History", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11008v1", "sha256": "fc8a88d4b380ac466b81452350f508615a76ef4157a011373be41ba3e84423d5", "local_html_sha256": "fc8a88d4b380ac466b81452350f508615a76ef4157a011373be41ba3e84423d5", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11008v1", "result": "retrieved", "sha256": "fc8a88d4b380ac466b81452350f508615a76ef4157a011373be41ba3e84423d5" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.11007", "title": "Curating Always-Loaded Context for LLM Agents: A Capacitated Assortment Model with Censored Feedback", "published": "2026-10-07T23:44:32Z", "reviews": [ { "id": "2610.11007", "title": "Curating Always-Loaded Context for LLM Agents: A Capacitated Assortment Model with Censored Feedback", "published": "2026-10-07T23:44:32Z", "decision": "reviewed_deferred", "reason": "已读capacitated context model/experiments：有限attention/token成本assortment与censoredfeedback，删除ignored规则会误删helpful，3200realcontext项改成可评分指令。下一批需真实session与外部compliance而非理论模型代替policy能力，本论文不授权改本仓AGENTS。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 MODEL", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 EXPERIMENTS", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.11007v1", "sha256": "032b07f4bf6c3b6a3ddaa5db5e37334966c33eb45175af9d436b0f33104679e0", "local_html_sha256": "032b07f4bf6c3b6a3ddaa5db5e37334966c33eb45175af9d436b0f33104679e0", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.11007v1", "result": "retrieved", "sha256": "032b07f4bf6c3b6a3ddaa5db5e37334966c33eb45175af9d436b0f33104679e0" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10990", "title": "Omni-Diffusion-Distill: Few-Step Distillation of Unified Multimodal Diffusion Large Language Models", "published": "2026-10-07T23:24:16Z", "reviews": [ { "id": "2610.10990", "title": "Omni-Diffusion-Distill: Few-Step Distillation of Unified Multimodal Diffusion Large Language Models", "published": "2026-10-07T23:24:16Z", "decision": "reviewed_deferred", "reason": "Meta unified离散dLLM teachertrajectory回放+studentrollout精炼，collision/entropyguidance；LuminaDiMOO8B17benchmark需真实两stage训练。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Method", "location": ".cache/oct12-scan/fulltext/2610.10990.json sections text prefix", "characters_read": [ 0, 1000 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.10990.json sections text prefix", "characters_read": [ 0, 1150 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 Method", "section_index": 2, "characters_read": [ 0, 1408 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "4 Experiments", "section_index": 3, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "3 Method", "location": "text after actual 3.2 Motivation / 3.3 trajectory replay", "characters_read": [ 0, 2200 ], "scope": "bounded excerpt; raw source available, cached section extraction may be capped" }, { "role": "additional targeted method/evaluation", "heading": "3 Method", "location": "text after actual 3.4 On-Policy Refinement", "characters_read": [ 0, 2200 ], "scope": "bounded excerpt; raw source available, cached section extraction may be capped" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 158, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10990v1", "sha256": "625fbd0f40eeafd65f8f636a2c057991dd292a94620b2a5aa82c4d18dd1a664d", "local_path": ".cache/oct12-scan/fulltext/2610.10990.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10990v1", "result": "retrieved", "sha256": "625fbd0f40eeafd65f8f636a2c057991dd292a94620b2a5aa82c4d18dd1a664d" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10990", "sha256": "9350a4e434299a8b3aa05b99a0cb4548634b7d392710003b227b87521bc36724", "local_path": ".cache/oct12-scan/fulltext/2610.10990.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 23:24:16 UTC (7,611 KB)", "published": "2026-10-07T23:24:16Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10988", "title": "Back in Style: A Sociolinguistic Approach to Authoring and Measuring Persona Fidelity in User Simulation", "published": "2026-10-07T23:21:49Z", "reviews": [ { "id": "2610.10988", "title": "Back in Style: A Sociolinguistic Approach to Authoring and Measuring Persona Fidelity in User Simulation", "published": "2026-10-07T23:21:49Z", "decision": "reviewed_deferred", "reason": "已读persona stylometry方法/结果：定量style schema vs archetype controlledA/B、fixedagent WebRTC互动，5models distinguishability/adherence。A/B是user simulator pilot非生产traffic，style fidelity不等于naturalness/代表真实用户；下一批评测infra。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10988v1", "sha256": "6f3d58bbfcebf4240f537c809497de82fc3e856954347c3cac2bf4b0164ef643", "local_html_sha256": "6f3d58bbfcebf4240f537c809497de82fc3e856954347c3cac2bf4b0164ef643", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10988v1", "result": "retrieved", "sha256": "6f3d58bbfcebf4240f537c809497de82fc3e856954347c3cac2bf4b0164ef643" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10984", "title": "Fluid-Gen-Zero: Grounding Pretrained Video Generators in Physics without Training", "published": "2026-10-07T23:16:32Z", "reviews": [ { "id": "2610.10984", "title": "Fluid-Gen-Zero: Grounding Pretrained Video Generators in Physics without Training", "published": "2026-10-07T23:16:32Z", "decision": "reviewed_deferred", "reason": "Meta协作流体simulator+VLMclipplan+regionlatentguidance，43scene三video骨干/13人人评；code待release，真实simulator与GPU待实施。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Method", "location": ".cache/oct12-scan/fulltext/2610.10984.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.10984.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 Method", "section_index": 2, "characters_read": [ 0, 1402 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "2 Generation-Time Planning Details", "section_index": 6, "characters_read": [ 0, 1402 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "4 Implementation Details", "section_index": 8, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "7 User Study Details", "section_index": 11, "characters_read": [ 0, 1800 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 159, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10984v1", "sha256": "7e3d9d9181be1cf71115f5afc8e52e66fe81d92fa176c82c28ec3e5566247cb9", "local_path": ".cache/oct12-scan/fulltext/2610.10984.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10984v1", "result": "retrieved", "sha256": "7e3d9d9181be1cf71115f5afc8e52e66fe81d92fa176c82c28ec3e5566247cb9" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10984", "sha256": "dc2408c751afdca1f28646878fb7839144cc1d46f9391db222ba56747692b6d3", "local_path": ".cache/oct12-scan/fulltext/2610.10984.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 23:16:32 UTC (10,404 KB)", "published": "2026-10-07T23:16:32Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10974", "title": "Budgeted Multi-Source Counterfactual Annotation for Off-Policy Evaluation", "published": "2026-10-07T22:55:15Z", "reviews": [ { "id": "2610.10974", "title": "Budgeted Multi-Source Counterfactual Annotation for Off-Policy Evaluation", "published": "2026-10-07T22:55:15Z", "decision": "rejected_or_out_of_scope", "reason": "多源counterfactual annotation成本分配与MM/DP主要contextual-bandit OPE，LLM只是教育annotation source；读formulation/numerical experiments，无LLM agent后训练核心，不纳入冻结实施。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Problem Formulation", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "7 Numerical Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10974v1", "sha256": "2cda31d53083bb7e488dd83b1a35d46b3dd943b18946e5b94bed772151340832", "local_html_sha256": "2cda31d53083bb7e488dd83b1a35d46b3dd943b18946e5b94bed772151340832", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10974v1", "result": "retrieved", "sha256": "2cda31d53083bb7e488dd83b1a35d46b3dd943b18946e5b94bed772151340832" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10962", "title": "iAm.md: Robot Skill Self-Assessment through Agentic Introspection for Unknown Open-Vocabulary Domains", "published": "2026-10-07T22:28:28Z", "reviews": [ { "id": "2610.10962", "title": "iAm.md: Robot Skill Self-Assessment through Agentic Introspection for Unknown Open-Vocabulary Domains", "published": "2026-10-07T22:28:28Z", "decision": "reviewed_deferred", "reason": "已读iAm.md方法/semanticmap结果：机器人persistent evidence、VLM+segmentation object records、ROS2 code生成/执行，在模拟TIAGo任务。下一批需真实感知模块与模拟执行，Markdown self-description不能当实际skill success。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experimental Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10962v1", "sha256": "99f9e0e3a14bd729480dd66b2af73c09c597d31f8a870c381ce67d8aba0a525c", "local_html_sha256": "99f9e0e3a14bd729480dd66b2af73c09c597d31f8a870c381ce67d8aba0a525c", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10962v1", "result": "retrieved", "sha256": "99f9e0e3a14bd729480dd66b2af73c09c597d31f8a870c381ce67d8aba0a525c" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10961", "title": "Cross-Provider Review as a Runtime Contract for Coding Agents: A Controlled Pilot and Fault-Injection Study", "published": "2026-10-07T22:28:22Z", "reviews": [ { "id": "2610.10961", "title": "Cross-Provider Review as a Runtime Contract for Coding Agents: A Controlled Pilot and Fault-Injection Study", "published": "2026-10-07T22:28:22Z", "decision": "reviewed_deferred", "reason": "已读cross-provider pilot/影子study：完整input、restrictedreviewer、boundedprocess/exitstatus与每attempt durableevidence；20paired task为agent自造，25formal因exit-status缺陷停用validcohort、重新从零。下一批infra应保留负结果，不把修复后的diagnostic当现场可靠性或已过gate。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Controlled Pilot: Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "7 Prospective Shadow Study: Review Allocation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10961v1", "sha256": "cc44b70da2cec72bf2b6a1c2172e2d21e36448de9c3865301db6784a3da7b381", "local_html_sha256": "cc44b70da2cec72bf2b6a1c2172e2d21e36448de9c3865301db6784a3da7b381", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10961v1", "result": "retrieved", "sha256": "cc44b70da2cec72bf2b6a1c2172e2d21e36448de9c3865301db6784a3da7b381" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10952", "title": "SPD-MetaFormer is what you need for small-data brain decoding", "published": "2026-10-07T22:09:32Z", "reviews": [ { "id": "2610.10952", "title": "SPD-MetaFormer is what you need for small-data brain decoding", "published": "2026-10-07T22:09:32Z", "decision": "rejected_or_out_of_scope", "reason": "协方差SPD/Fréchet小数据EEG脑解码架构，不是本批LLM或多模态foundation后训练算法。", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "read_scope": "abstract_scope_exclusion; original HTML transport is not counted as reading", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 160, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10952v1", "sha256": "c4897e57a4c91f13d20cdecbfb5c733863bdb453750c7b44eb5b5d8cfefa201d", "local_path": ".cache/oct12-scan/fulltext/2610.10952.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10952v1", "result": "retrieved", "sha256": "c4897e57a4c91f13d20cdecbfb5c733863bdb453750c7b44eb5b5d8cfefa201d" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10952", "sha256": "dd4348e5c9bd2dbd6efabb6d424de4ac6c4a916b9b9e1443d942c5d6a0be103f", "local_path": ".cache/oct12-scan/fulltext/2610.10952.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 22:09:32 UTC (378 KB)", "published": "2026-10-07T22:09:32Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10947", "title": "RH-Detect: A Unified Benchmark for Reward Hacking Detection", "published": "2026-10-07T21:52:33Z", "reviews": [ { "id": "2610.10947", "title": "RH-Detect: A Unified Benchmark for Reward Hacking Detection", "published": "2026-10-07T21:52:33Z", "decision": "reviewed_deferred", "reason": "RH-Detect统一11 sources schema并source-heldout detectorSFT/GRPO，pooled AUROC不能掩multi-turn跌落；下一批evaluation integrity需保originalprovenance/labels且heldoutsource选参隔离。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Dataset Construction", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Zero-Shot Detector Evaluation", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10947v1", "sha256": "e4700bbc349d485cefb8e96053c2971d567f6cd51bd6d35dda0ee3a0276f4bac", "local_html_sha256": "e4700bbc349d485cefb8e96053c2971d567f6cd51bd6d35dda0ee3a0276f4bac", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10947v1", "result": "retrieved", "sha256": "e4700bbc349d485cefb8e96053c2971d567f6cd51bd6d35dda0ee3a0276f4bac" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10942", "title": "StoreBench: A Live-Commerce Environment for Evaluating and Training Autonomous Operator Agents", "published": "2026-10-07T21:46:10Z", "reviews": [ { "id": "2610.10942", "title": "StoreBench: A Live-Commerce Environment for Evaluating and Training Autonomous Operator Agents", "published": "2026-10-07T21:46:10Z", "decision": "reviewed_deferred", "reason": "StoreBench Medusa live-commerce动态环境、action-driven时间、anchor-calibrated reward和GRPO训练有heldout split；完整环境/eval被保留未公开，不能承诺完整复现，下一批需确认公开sample与scorer可执行范围。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 The StoreBench environment", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experimental setup", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10942v1", "sha256": "ff478d4dbdf2da211ee9d046ef0091a4ddd7933e7200be38d65cd84fc7382ea7", "local_html_sha256": "ff478d4dbdf2da211ee9d046ef0091a4ddd7933e7200be38d65cd84fc7382ea7", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10942v1", "result": "retrieved", "sha256": "ff478d4dbdf2da211ee9d046ef0091a4ddd7933e7200be38d65cd84fc7382ea7" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10932", "title": "World-Model Policy Arbiter for Goal-Conditioned Reinforcement Learning", "published": "2026-10-07T21:41:33Z", "reviews": [ { "id": "2610.10932", "title": "World-Model Policy Arbiter for Goal-Conditioned Reinforcement Learning", "published": "2026-10-07T21:41:33Z", "decision": "rejected_or_out_of_scope", "reason": "WMPA learned statespaceworldmodel arbitrate frozen goal-conditioned continuouscontrolpolicies，18 OGBench state datasets；不是LLM/foundation model后训练，portfolio/value机制可参考但当前paperimplementation outofscope。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method: World-Model Policy Arbiter", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10932v1", "sha256": "e4db227678e0cbc1048b6f1b10ff5fba0e6612f0bc9eb91b377e1cd5ef35add6", "local_html_sha256": "e4db227678e0cbc1048b6f1b10ff5fba0e6612f0bc9eb91b377e1cd5ef35add6", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10932v1", "result": "retrieved", "sha256": "e4db227678e0cbc1048b6f1b10ff5fba0e6612f0bc9eb91b377e1cd5ef35add6" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10926", "title": "Adaptive Multi-Discriminator WGAN Framework for Resource-Constrained Internet of Vehicles Using Reinforcement Learning and Game Theory", "published": "2026-10-07T21:28:46Z", "reviews": [ { "id": "2610.10926", "title": "Adaptive Multi-Discriminator WGAN Framework for Resource-Constrained Internet of Vehicles Using Reinforcement Learning and Game Theory", "published": "2026-10-07T21:28:46Z", "decision": "rejected_or_out_of_scope", "reason": "已读MD-WGAN system/experiment：NS3车辆网、DQN选择discriminator和game epochalloc，NGSIM交通预测，是GAN资源调度非LLM agent；真实轨迹数据不等于真实部署实验。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "III System Model", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "V Experiments and Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10926v1", "sha256": "7288706c991b4fb4c8d2acb19ce779c0442cb9e4ddf81a25e6861192ebc81b74", "local_html_sha256": "7288706c991b4fb4c8d2acb19ce779c0442cb9e4ddf81a25e6861192ebc81b74", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10926v1", "result": "retrieved", "sha256": "7288706c991b4fb4c8d2acb19ce779c0442cb9e4ddf81a25e6861192ebc81b74" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10920", "title": "Language Models for Page-Level Layout Decisions in E-commerce Search", "published": "2026-10-07T21:19:47Z", "reviews": [ { "id": "2610.10920", "title": "Language Models for Page-Level Layout Decisions in E-commerce Search", "published": "2026-10-07T21:19:47Z", "decision": "rejected_or_out_of_scope", "reason": "Walmart页面布局工作实际研究offline LM evaluator；A/B出现于动机/评价成本和引用前作，不是该方法新生产实验。按正文完整评测范围拒绝工业门槛，非摘要关键词过滤。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "bounded_methods_and_evaluation", "review_artifact": "research-audits/oct12-review-closure/recommendation-reviews.json", "evidence_locations": [ "§1 Introduction", "§3–§5 offline evaluation" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10920v1", "sha256": "99d2e214224503992140a7ccab4187a879b216fc3c502b086b5041f5f704930e" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "recommendation" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10918", "title": "Large Language Models for Machine Translation Quality Annotation: Humans and Models Are Both Challenged", "published": "2026-10-07T21:17:26Z", "reviews": [ { "id": "2610.10918", "title": "Large Language Models for Machine Translation Quality Annotation: Humans and Models Are Both Challenged", "published": "2026-10-07T21:17:26Z", "decision": "reviewed_deferred", "reason": "Apple MT MQM/ESA score/span人机agreement；publicWMT23/25+内部70语言，agreement多低于可靠阈值，内部数据不可冒称公开可复现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "3 Experimental Setup", "location": ".cache/oct12-scan/fulltext/2610.10918.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "4 Results", "section_index": 3, "characters_read": [ 0, 1600 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 161, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10918v1", "sha256": "b4546249d382af74718346512459b930a3cd372bafa96be6a132eff1c78b912d", "local_path": ".cache/oct12-scan/fulltext/2610.10918.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10918v1", "result": "retrieved", "sha256": "b4546249d382af74718346512459b930a3cd372bafa96be6a132eff1c78b912d" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10918", "sha256": "748bdc3ce9c7a1bb1867b5dbeccec2fb08adcbd13059afb2c4337d313b87d255", "local_path": ".cache/oct12-scan/fulltext/2610.10918.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 21:17:26 UTC (1,024 KB)", "published": "2026-10-07T21:17:26Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10912", "title": "When Listening Becomes Easier: Scrubbing Visual Cues for Shortcut-Free VLAs", "published": "2026-10-07T21:11:23Z", "reviews": [ { "id": "2610.10912", "title": "When Listening Becomes Easier: Scrubbing Visual Cues for Shortcut-Free VLAs", "published": "2026-10-07T21:11:23Z", "decision": "reviewed_deferred", "reason": "VLA visualtaskadversary scrubbing与actionmargin分析，LIBERO三seed/realrobot；真实policyrollout和OODcue控制待下一批。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "IV Task Scrubbing", "location": ".cache/oct12-scan/fulltext/2610.10912.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "V Results", "location": ".cache/oct12-scan/fulltext/2610.10912.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 162, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10912v1", "sha256": "2992947d625867b5789ab6f633a675508a1c2db643c5f7cde995eb9ce60119b2", "local_path": ".cache/oct12-scan/fulltext/2610.10912.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10912v1", "result": "retrieved", "sha256": "2992947d625867b5789ab6f633a675508a1c2db643c5f7cde995eb9ce60119b2" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10912", "sha256": "0685fda936a40eaa637c09e03618248d0c5e5db297499364c6c9b9d295344012", "local_path": ".cache/oct12-scan/fulltext/2610.10912.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 21:11:23 UTC (2,673 KB)", "published": "2026-10-07T21:11:23Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10906", "title": "Reading the Room: Foundations, Design, and Challenges of Normative Competence in LLMs", "published": "2026-10-07T21:02:52Z", "reviews": [ { "id": "2610.10906", "title": "Reading the Room: Foundations, Design, and Challenges of Normative Competence in LLMs", "published": "2026-10-07T21:02:52Z", "decision": "reviewed_deferred", "reason": "已读normative competence设定/实验：synthetic sillynorm社区debate隔离prior，normmodule对不同norm/无关idiosyncrasy的消融与负结果。下一批行为infra需观察真实interaction学习，不能用hiddennorm答案或simple imitation实现norm推断。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "5 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Evaluating Normative Competence via Community Debate", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10906v1", "sha256": "05b9ce3abc6c26502a9432d9fbb7694508a7859bcfac8924feef1e9eaf8e3002", "local_html_sha256": "05b9ce3abc6c26502a9432d9fbb7694508a7859bcfac8924feef1e9eaf8e3002", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10906v1", "result": "retrieved", "sha256": "05b9ce3abc6c26502a9432d9fbb7694508a7859bcfac8924feef1e9eaf8e3002" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10893", "title": "Less from More: Reinforcing Sparse Video Reasoning from Dense References", "published": "2026-10-07T20:51:59Z", "reviews": [ { "id": "2610.10893", "title": "Less from More: Reinforcing Sparse Video Reasoning from Dense References", "published": "2026-10-07T20:51:59Z", "decision": "reviewed_deferred", "reason": "SAVER denseview仅trainreference、1250groundingRL可靠门控reward，稀疏test不预扫描；三grounding/六videoQA真实policy待实施。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "Method", "location": ".cache/oct12-scan/fulltext/2610.10893.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "Experiments", "location": ".cache/oct12-scan/fulltext/2610.10893.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 163, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10893v1", "sha256": "3d1ea833e016c7c5aa8f871819b8acba8424c55e489ac1039e61a95182de7137", "local_path": ".cache/oct12-scan/fulltext/2610.10893.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10893v1", "result": "retrieved", "sha256": "3d1ea833e016c7c5aa8f871819b8acba8424c55e489ac1039e61a95182de7137" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10893", "sha256": "0b63448b291e85a42de5fda7c148f56dcd664b6df873954a14aa246a24dd2b21", "local_path": ".cache/oct12-scan/fulltext/2610.10893.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 20:51:59 UTC (30,525 KB)", "published": "2026-10-07T20:51:59Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10889", "title": "SPLIT-RL: Staged Perception-Language Reasoning Training with Claim-Level Advantages", "published": "2026-10-07T20:42:54Z", "reviews": [ { "id": "2610.10889", "title": "SPLIT-RL: Staged Perception-Language Reasoning Training with Claim-Level Advantages", "published": "2026-10-07T20:42:54Z", "decision": "reviewed_deferred", "reason": "SPLIT VR/LR分phase单weight，atomicvisualclaims按type分组优势，matchedsteps/outputbudget；oracle诊断不进推理接口。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.10889.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 SPLIT-RL", "section_index": 2, "characters_read": [ 0, 1600 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 164, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10889v1", "sha256": "e23fba5a91b72b80fac9eee015a0544c115fcaab02fd33544da891f1130d747f", "local_path": ".cache/oct12-scan/fulltext/2610.10889.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10889v1", "result": "retrieved", "sha256": "e23fba5a91b72b80fac9eee015a0544c115fcaab02fd33544da891f1130d747f" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10889", "sha256": "73dcbae5b477fee7924b8a33ce9258757fcb11c5dc77e47322213cdb4f4ba4b0", "local_path": ".cache/oct12-scan/fulltext/2610.10889.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 20:42:54 UTC (1,570 KB)", "published": "2026-10-07T20:42:54Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10878", "title": "Stochastic Teacher Intervention for Agentic On-Policy Distillation", "published": "2026-10-07T20:29:29Z", "reviews": [ { "id": "2610.10878", "title": "Stochastic Teacher Intervention for Agentic On-Policy Distillation", "published": "2026-10-07T20:29:29Z", "decision": "reviewed_deferred", "reason": "STI-OPD用k3师生KL映射随机teacher intervention，mixed policy token须importance-weight reverse KL校正；工具推理/ALFWorld/ScienceWorld多benchmark，下一批必须实际环境执行teacher替换，不能仅loss proxy。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10878v1", "sha256": "a7bd7272519466f5a2080b196019a07433b2bcc9d482616930f25be4105a88fd", "local_html_sha256": "a7bd7272519466f5a2080b196019a07433b2bcc9d482616930f25be4105a88fd", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10878v1", "result": "retrieved", "sha256": "a7bd7272519466f5a2080b196019a07433b2bcc9d482616930f25be4105a88fd" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10868", "title": "Conversational Voice Aesthetic Model with Reinforcement Learning from Human Listeners", "published": "2026-10-07T20:13:03Z", "reviews": [ { "id": "2610.10868", "title": "Conversational Voice Aesthetic Model with Reinforcement Learning from Human Listeners", "published": "2026-10-07T20:13:03Z", "decision": "reviewed_deferred", "reason": "CVAM Qwen2.5-Omni voiceaesthetics SFT+humanlabelGRPO，objective pitch/pacing与subjective emotion/delivery区分，CANDOR human majority评测；下一批需真实音频与多listener训练reward，不把singlerater当确定gold。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "2 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Results", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10868v1", "sha256": "f8c082d200d9d6a87d479fc573d2846122f42869a9dcaa1845bb14963100324c", "local_html_sha256": "f8c082d200d9d6a87d479fc573d2846122f42869a9dcaa1845bb14963100324c", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10868v1", "result": "retrieved", "sha256": "f8c082d200d9d6a87d479fc573d2846122f42869a9dcaa1845bb14963100324c" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10858", "title": "RFChipAgent: Multi-Agentic AI Flow for Analog/RF Chip Design", "published": "2026-10-07T20:02:33Z", "reviews": [ { "id": "2610.10858", "title": "RFChipAgent: Multi-Agentic AI Flow for Analog/RF Chip Design", "published": "2026-10-07T20:02:33Z", "decision": "reviewed_deferred", "reason": "Qualcomm RFIC superagent组织private document FAISS、topology和schematic/testbench agent、TPE+CMA-ES simulator loop与trust-scored simulationDB。; GF22FDSOI 60GHz two-stage LNA一个family的设计campaign；成本/时长为系统run描述。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.10858v1#S3", "https://arxiv.org/html/2610.10858v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10858v1", "sha256": "30b48143443f57b1b1af428da06da2eeda556007e7f17cda0fcb806e720ee3c1" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10854", "title": "KDFP: A first-principles approach to knowledge distillation in large language models", "published": "2026-10-07T19:59:07Z", "reviews": [ { "id": "2610.10854", "title": "KDFP: A first-principles approach to knowledge distillation in large language models", "published": "2026-10-07T19:59:07Z", "decision": "reviewed_deferred", "reason": "KDFP whiteboxgeneralknowledgeKD在continualpretrainPythia和9tasks对照、ephemeralparameterreduction提升训练效率；foundationKD下一批需查actual目标/参数临时冻结恢复，不能与instructionOPD混称。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "5 Methodology", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "6 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10854v1", "sha256": "5c486aafb0efc59620272162a5403161289ee9e5ecabd1591d625e9631877548", "local_html_sha256": "5c486aafb0efc59620272162a5403161289ee9e5ecabd1591d625e9631877548", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10854v1", "result": "retrieved", "sha256": "5c486aafb0efc59620272162a5403161289ee9e5ecabd1591d625e9631877548" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10845", "title": "Real Long-Term Memory for AI: A 50-Million-Token Window That Is Faster and Cheaper Than Recompute", "published": "2026-10-07T19:50:53Z", "reviews": [ { "id": "2610.10845", "title": "Real Long-Term Memory for AI: A 50-Million-Token Window That Is Faster and Cheaper Than Recompute", "published": "2026-10-07T19:50:53Z", "decision": "reviewed_deferred", "reason": "galahadKV encryptedNVMe16kblock restore/vLLM50Mstream；是单blockKV复用非50M共同attention，许可/terabytecost/独立probe需核。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "5  Experimental setup", "location": ".cache/oct12-scan/fulltext/2610.10845.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3  System under test", "section_index": 2, "characters_read": [ 0, 796 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "4  Anti-cheat protocol", "section_index": 3, "characters_read": [ 0, 1600 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "6  Results", "section_index": 5, "characters_read": [ 0, 1600 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 165, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10845v1", "sha256": "f0f5e57d5bf6e16d26e6db161665007d83a1d7c7b400f24f8e0b68f23ffe785b", "local_path": ".cache/oct12-scan/fulltext/2610.10845.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10845v1", "result": "retrieved", "sha256": "f0f5e57d5bf6e16d26e6db161665007d83a1d7c7b400f24f8e0b68f23ffe785b" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10845", "sha256": "c45484c8bf9826d6b47926deae6acfaef3b8a5e45cea86ed684d53a049b6a72b", "local_path": ".cache/oct12-scan/fulltext/2610.10845.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 19:50:53 UTC (11 KB)", "published": "2026-10-07T19:50:53Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.12517", "title": "TerraAgent: Demonstration-Grounded Multi-Agent Reasoning for Long-Context Remote Sensing Visual Question Answering", "published": "2026-10-07T19:50:09Z", "reviews": [ { "id": "2610.12517", "title": "TerraAgent: Demonstration-Grounded Multi-Agent Reasoning for Long-Context Remote Sensing Visual Question Answering", "published": "2026-10-07T19:50:09Z", "decision": "reviewed_deferred", "reason": "TerraAgent/AVA GSDdemo检索、工具DAGledger和counterfactualevidence审查；publicRSVQA/VRS/LHRS需demo与test隔离、真图像干预。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Method", "location": ".cache/oct12-scan/fulltext/2610.12517.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experimental Setup", "location": ".cache/oct12-scan/fulltext/2610.12517.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 166, "primary_fulltext": { "url": "https://arxiv.org/html/2610.12517v1", "sha256": "5cc38b10ccba9dde888ac96d648076a93d5c93ffffa858ca7204d764b1945529", "local_path": ".cache/oct12-scan/fulltext/2610.12517.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12517v1", "result": "retrieved", "sha256": "5cc38b10ccba9dde888ac96d648076a93d5c93ffffa858ca7204d764b1945529" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.12517", "sha256": "cf89f1979330b05887cfe9f05d664bedf8e5cf6d478703696a539b820ab662f6", "local_path": ".cache/oct12-scan/fulltext/2610.12517.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 19:50:09 UTC (1,317 KB)", "published": "2026-10-07T19:50:09Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10833", "title": "On the Clock: Towards Punctual and Productive Time-Budgeted AI Agents", "published": "2026-10-07T19:35:08Z", "reviews": [ { "id": "2610.10833", "title": "On the Clock: Towards Punctual and Productive Time-Budgeted AI Agents", "published": "2026-10-07T19:35:08Z", "decision": "reviewed_deferred", "reason": "已读timebudget setup/results：harness clock+deadline enforcement与GRPO budgetreward，MLE5task/Zork未知预算transfer；准时与productive quality分开，RL可能重复动作/shortbudget collapse。下一批保留无task提升负结果，不能靠sleep/stop规则宣称策略学会时间分配。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10833v1", "sha256": "e7ae99edeace950f9861f6daa71265cd47fe7706e9f34bc8ebfbc3cc78e709ff", "local_html_sha256": "e7ae99edeace950f9861f6daa71265cd47fe7706e9f34bc8ebfbc3cc78e709ff", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10833v1", "result": "retrieved", "sha256": "e7ae99edeace950f9861f6daa71265cd47fe7706e9f34bc8ebfbc3cc78e709ff" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.12516", "title": "TraceICL: Thread-Resolved Demonstration Curation with Consistency Calibration for Robust Multimodal In-Context Learning", "published": "2026-10-07T19:31:54Z", "reviews": [ { "id": "2610.12516", "title": "TraceICL: Thread-Resolved Demonstration Curation with Consistency Calibration for Robust Multimodal In-Context Learning", "published": "2026-10-07T19:31:54Z", "decision": "reviewed_deferred", "reason": "TraceICL threadbind复原、marginalinfluence选demo、弱模型confidence解码校准；frozen4shot/噪声bench需真实forward与split。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Method", "location": ".cache/oct12-scan/fulltext/2610.12516.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.12516.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 167, "primary_fulltext": { "url": "https://arxiv.org/html/2610.12516v1", "sha256": "55fcd6173a6122a18994b896538c2c241ecbfdb72434f36f7a6fcfda0c439750", "local_path": ".cache/oct12-scan/fulltext/2610.12516.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.12516v1", "result": "retrieved", "sha256": "55fcd6173a6122a18994b896538c2c241ecbfdb72434f36f7a6fcfda0c439750" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.12516", "sha256": "25a1d5eeb2ad66d83b620b02f9aed2001d889f749dc447649b9820f581af1dde", "local_path": ".cache/oct12-scan/fulltext/2610.12516.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 19:31:54 UTC (1,363 KB)", "published": "2026-10-07T19:31:54Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10787", "title": "NavGPT-3: Harnessing Context in a Hierarchical Navigation Runtime", "published": "2026-10-07T18:45:56Z", "reviews": [ { "id": "2610.10787", "title": "NavGPT-3: Harnessing Context in a Hierarchical Navigation Runtime", "published": "2026-10-07T18:45:56Z", "decision": "reviewed_deferred", "reason": "NavGPT-3分离reasoning/acting/monitoring线程，VLA codec视觉token分配及embodied harness有R2R-CE/RxR-CE及真robot实验；主要embodied foundation/agent系统，下一批需模型工具权限与动作runtime，非本批posttraining operator。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method: Harness, Runtime, and VLA Co-Design", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10787v1", "sha256": "14eb5d50fb5286992716de3db7a06b598df7c7c00e98f3cb9e28f829a51dc404", "local_html_sha256": "14eb5d50fb5286992716de3db7a06b598df7c7c00e98f3cb9e28f829a51dc404", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10787v1", "result": "retrieved", "sha256": "14eb5d50fb5286992716de3db7a06b598df7c7c00e98f3cb9e28f829a51dc404" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10786", "title": "Plan-and-Patch: Diffusion Language Models for Agentic Planning", "published": "2026-10-07T18:45:45Z", "reviews": [ { "id": "2610.10786", "title": "Plan-and-Patch: Diffusion Language Models for Agentic Planning", "published": "2026-10-07T18:45:45Z", "decision": "reviewed_deferred", "reason": "已读Plan-and-Patch方法/实验：真实diffusion planner parallelunmask、region infill保prefix/suffix，与独立executor失败region选择，NaturalPlan/ALFWorld/TextCraft训练。下一批须dLLM与联合generation/repair训练，AR字符串编辑非定义机制。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 The Plan-and-Patch Framework", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10786v1", "sha256": "4528d3a88b80a9357f7358fa7fddfc6c1f212b7e8085fdd2faa46e739fef595c", "local_html_sha256": "4528d3a88b80a9357f7358fa7fddfc6c1f212b7e8085fdd2faa46e739fef595c", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10786v1", "result": "retrieved", "sha256": "4528d3a88b80a9357f7358fa7fddfc6c1f212b7e8085fdd2faa46e739fef595c" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10782", "title": "VICO: Visual Environments Co-Evolving for Vision-Language Model Reasoning", "published": "2026-10-07T18:39:32Z", "reviews": [ { "id": "2610.10782", "title": "VICO: Visual Environments Co-Evolving for Vision-Language Model Reasoning", "published": "2026-10-07T18:39:32Z", "decision": "reviewed_deferred", "reason": "VICO actor/EnvRewriter联合进化可验证image-side结构并renderer重画，passrate能力边界reward；9bench真实双policy未实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Methodology", "location": ".cache/oct12-scan/fulltext/2610.10782.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5 Experiments", "location": ".cache/oct12-scan/fulltext/2610.10782.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 168, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10782v1", "sha256": "eab4538b1783ca4f8d5223bf1cbf8a8790d9f4ca974ab7d130b3e124abda0cf1", "local_path": ".cache/oct12-scan/fulltext/2610.10782.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10782v1", "result": "retrieved", "sha256": "eab4538b1783ca4f8d5223bf1cbf8a8790d9f4ca974ab7d130b3e124abda0cf1" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10782", "sha256": "94afa596fe8d792096a1b87eb5da6d1d4d8c372fbdc855b763d15cc22d57fc1c", "local_path": ".cache/oct12-scan/fulltext/2610.10782.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 18:39:32 UTC (22,719 KB)", "published": "2026-10-07T18:39:32Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10768", "title": "Strategic Investment Decision Making for Value Creation in Energy Transition: A Reinforcement Learning Approach", "published": "2026-10-07T18:29:31Z", "reviews": [ { "id": "2610.10768", "title": "Strategic Investment Decision Making for Value Creation in Energy Transition: A Reinforcement Learning Approach", "published": "2026-10-07T18:29:31Z", "decision": "rejected_or_out_of_scope", "reason": "已读摘要，官方正文未取得：2050能源投资simulation中传统RL多objective sequential策略，非LLM agent，本批范围外；仅模型假设结果非财务建议或真实收益证据。", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "read_scope": "abstract_only_primary_fulltext_unavailable", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "url": "https://arxiv.org/abs/2610.10768", "sha256": null, "local_html_sha256": null, "local_html_sha256_matches": false, "attempts": [ { "url": "https://arxiv.org/html/2610.10768v1", "result": "failed", "error": "404 Client Error: Not Found for url: https://arxiv.org/html/2610.10768v1" }, { "url": "https://arxiv.org/html/2610.10768v1", "result": "failed", "error": "404 Client Error: Not Found for url: https://arxiv.org/html/2610.10768v1" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10760", "title": "DOGS: Design-Space Sampling for Prompt-Driven Logo Generation", "published": "2026-10-07T18:24:11Z", "reviews": [ { "id": "2610.10760", "title": "DOGS: Design-Space Sampling for Prompt-Driven Logo Generation", "published": "2026-10-07T18:24:11Z", "decision": "reviewed_deferred", "reason": "DOGS typedlogo图grammar+GFlowNet originalityreward，L3Dhashtrainval与两同T2Irenderer；语料近似原创性不是法律侵权结论。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Methodology", "location": ".cache/oct12-scan/fulltext/2610.10760.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.10760.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 169, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10760v1", "sha256": "a4ba9cc8ef18dc0a683411e2bcb11272730b817e5b9c70cdda480726d671435d", "local_path": ".cache/oct12-scan/fulltext/2610.10760.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10760v1", "result": "retrieved", "sha256": "a4ba9cc8ef18dc0a683411e2bcb11272730b817e5b9c70cdda480726d671435d" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10760", "sha256": "0e4f675851dc5dd716d3ef803ab0c17aa48e1348f63f3f0855c4bdca67bb50cf", "local_path": ".cache/oct12-scan/fulltext/2610.10760.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 18:24:11 UTC (9,156 KB)", "published": "2026-10-07T18:24:11Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10740", "title": "Conversational Task Disambiguation over Tabular Data: Leakage-Aware Formulation, Benchmark Suite, and Training", "published": "2026-10-07T18:09:55Z", "reviews": [ { "id": "2610.10740", "title": "Conversational Task Disambiguation over Tabular Data: Leakage-Aware Formulation, Benchmark Suite, and Training", "published": "2026-10-07T18:09:55Z", "decision": "reviewed_deferred", "reason": "已读AmbiTab theory/benchmark：asking/solution policy分离、oracle/verifier显式访问边界、6ambiguousSQLdatasets及askerRL，judge-free leakage诊断。下一批必须限定oracle响应可泄漏内容，task success不混同clarification能力。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 AmbiTab benchmark suite", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 Theory", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10740v1", "sha256": "7b86cd394a8cd340589c96efbbbeb4319952ad0de39bc54064855f8ce869c8e0", "local_html_sha256": "7b86cd394a8cd340589c96efbbbeb4319952ad0de39bc54064855f8ce869c8e0", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10740v1", "result": "retrieved", "sha256": "7b86cd394a8cd340589c96efbbbeb4319952ad0de39bc54064855f8ce869c8e0" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10738", "title": "Lossy Compressive Text Autoencoders", "published": "2026-10-07T18:08:40Z", "reviews": [ { "id": "2610.10738", "title": "Lossy Compressive Text Autoencoders", "published": "2026-10-07T18:08:40Z", "decision": "reviewed_deferred", "reason": "textAE时间维压缩/离散残差bottleneck，Nemotron真实训练与语义/QA重建；lossy2.24bpb不能直接替代lossless语义完整性。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Method", "location": ".cache/oct12-scan/fulltext/2610.10738.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Evaluation", "location": ".cache/oct12-scan/fulltext/2610.10738.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 170, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10738v1", "sha256": "d1ca5b34349e4a22cab70ffaf8f085560a39262dd9b2d52f8b3c8a93402e81b3", "local_path": ".cache/oct12-scan/fulltext/2610.10738.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10738v1", "result": "retrieved", "sha256": "d1ca5b34349e4a22cab70ffaf8f085560a39262dd9b2d52f8b3c8a93402e81b3" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10738", "sha256": "759816c4d19184d805a88582dec38b7726e647225b348c855df0753165c9f49b", "local_path": ".cache/oct12-scan/fulltext/2610.10738.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 18:08:40 UTC (14,137 KB)", "published": "2026-10-07T18:08:40Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10697", "title": "On linearity or non-linearity in machine learning for quantum chaotic dynamics", "published": "2026-10-07T18:00:06Z", "reviews": [ { "id": "2610.10697", "title": "On linearity or non-linearity in machine learning for quantum chaotic dynamics", "published": "2026-10-07T18:00:06Z", "decision": "rejected_or_out_of_scope", "reason": "PXP量子少qubit混沌时间序列Transformer/DLinear预测，非LLM/foundation核心实现。", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "read_scope": "abstract_scope_exclusion; original HTML transport is not counted as reading", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 171, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10697v1", "sha256": "a5d38335887145ae063bd5fde27175a464f85f16ccf39ed0de6f0cba99f16b34", "local_path": ".cache/oct12-scan/fulltext/2610.10697.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10697v1", "result": "retrieved", "sha256": "a5d38335887145ae063bd5fde27175a464f85f16ccf39ed0de6f0cba99f16b34" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10697", "sha256": "44a601db416c58733214ccbf46a56013f276a98a010370c3a1883e4ea8cae04e", "local_path": ".cache/oct12-scan/fulltext/2610.10697.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 18:00:06 UTC (1,773 KB)", "published": "2026-10-07T18:00:06Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10536", "title": "Decoupling Exploration from Optimization in RLVR", "published": "2026-10-07T17:59:26Z", "reviews": [ { "id": "2610.10536", "title": "Decoupling Exploration from Optimization in RLVR", "published": "2026-10-07T17:59:26Z", "decision": "reviewed_deferred", "reason": "ExpDis separate explorer RND noveltyRL→correctness/qualityfilter→studentdistill→noNoveltyRL多轮，math同wallclockDAPO和passk比较；下一批需真实separatepolicies/RND和过滤成本，不能常量noveltyreward。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experimental Setup", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10536v1", "sha256": "22869b7a7c4ec36ef4ffd76c33fa33d0cfef6d29c16772a09b65b62d0b190d88", "local_html_sha256": "22869b7a7c4ec36ef4ffd76c33fa33d0cfef6d29c16772a09b65b62d0b190d88", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10536v1", "result": "retrieved", "sha256": "22869b7a7c4ec36ef4ffd76c33fa33d0cfef6d29c16772a09b65b62d0b190d88" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10533", "title": "EngramEdit: Decoupled Knowledge Updates in LLMs through Conditional Memory", "published": "2026-10-07T17:58:52Z", "reviews": [ { "id": "2610.10533", "title": "EngramEdit: Decoupled Knowledge Updates in LLMs through Conditional Memory", "published": "2026-10-07T17:58:52Z", "decision": "reviewed_deferred", "reason": "EngramEdit跨表达targetmemory联合sharedngram更新并reuse正则，CounterFact/ZsRE/MQuAKE；需真正conditionalmemory骨干/泛化保留测试。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Method", "location": ".cache/oct12-scan/fulltext/2610.10533.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.10533.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 172, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10533v1", "sha256": "c769ef72463657ad492c407251da3c1e30bd2fa955a1f2bc8a10f1a827cf9a91", "local_path": ".cache/oct12-scan/fulltext/2610.10533.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10533v1", "result": "retrieved", "sha256": "c769ef72463657ad492c407251da3c1e30bd2fa955a1f2bc8a10f1a827cf9a91" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10533", "sha256": "c394e5a15e1f6136ca154b01805023eb5aad902456456167e60547b5f7582b4a", "local_path": ".cache/oct12-scan/fulltext/2610.10533.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 17:58:52 UTC (726 KB)", "published": "2026-10-07T17:58:52Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10526", "title": "Rephrase Before You Act: Characterizing and Mitigating Language Sensitivity in Vision-Language-Action Models", "published": "2026-10-07T17:57:38Z", "reviews": [ { "id": "2610.10526", "title": "Rephrase Before You Act: Characterizing and Mitigating Language Sensitivity in Vision-Language-Action Models", "published": "2026-10-07T17:57:38Z", "decision": "reviewed_deferred", "reason": "VLA rephrasing从training任务phrase success提炼规则、deployment一次重写，无权重训练；SIMPLER/LIBERO heldout phrasing/tasked控制，转foundation/embodied robustness下一批，不能登记参数posttrain。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "V Reducing Language Sensitivity with LLM-Distilled Rephrasing Rules", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "VI Evaluating the Rulebooks", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10526v1", "sha256": "b6c7e43fe6dfa6ad65ede0685f45386bbdaa153094e68550b8b96bcafa4efe6c", "local_html_sha256": "b6c7e43fe6dfa6ad65ede0685f45386bbdaa153094e68550b8b96bcafa4efe6c", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10526v1", "result": "retrieved", "sha256": "b6c7e43fe6dfa6ad65ede0685f45386bbdaa153094e68550b8b96bcafa4efe6c" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10520", "title": "Distilling Graph Geometry: Knowledge Gap from GNNs to MLPs", "published": "2026-10-07T17:56:01Z", "reviews": [ { "id": "2610.10520", "title": "Distilling Graph Geometry: Knowledge Gap from GNNs to MLPs", "published": "2026-10-07T17:56:01Z", "decision": "rejected_or_out_of_scope", "reason": "曲率引导GNN到MLP的图节点/链接分类蒸馏，非本批foundation/LLM方法；prod加权指标不是生产实验。", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "read_scope": "abstract_scope_exclusion; original HTML transport is not counted as reading", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 173, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10520v1", "sha256": "9cf941a3b51a6516680d5f554c3f481536baa92c31613b16476ae9ecb1005e6b", "local_path": ".cache/oct12-scan/fulltext/2610.10520.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10520v1", "result": "retrieved", "sha256": "9cf941a3b51a6516680d5f554c3f481536baa92c31613b16476ae9ecb1005e6b" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10520", "sha256": "856326d2bed109526fed3abcd9037d7d00ba9cc9af38e126aaa61962f707d970", "local_path": ".cache/oct12-scan/fulltext/2610.10520.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 17:56:01 UTC (2,155 KB)", "published": "2026-10-07T17:56:01Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10507", "title": "RECAST: Learning to Compute the Right Context through Adaptive Evidence Routing", "published": "2026-10-07T17:51:08Z", "reviews": [ { "id": "2610.10507", "title": "RECAST: Learning to Compute the Right Context through Adaptive Evidence Routing", "published": "2026-10-07T17:51:08Z", "decision": "reviewed_deferred", "reason": "已读RECAST方法/实验：trainable RouterLM选primitive或custom operation，frozen CompilerLM编真实code、AnswerLM只见accepted evidence，SFT后GRPO六bench/3heldout。下一批需真实sandbox computation与router训练，不用固定tool选择或gold evidence。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Experimental Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 RECAST", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10507v1", "sha256": "94ae1fd3c8bb908ee36799d22fe9ff68d1f5068058c66db4d7a2c49bdbfe07f4", "local_html_sha256": "94ae1fd3c8bb908ee36799d22fe9ff68d1f5068058c66db4d7a2c49bdbfe07f4", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10507v1", "result": "retrieved", "sha256": "94ae1fd3c8bb908ee36799d22fe9ff68d1f5068058c66db4d7a2c49bdbfe07f4" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10491", "title": "Insights from Autoresearch for Solar Panel Segmentation", "published": "2026-10-07T17:45:31Z", "reviews": [ { "id": "2610.10491", "title": "Insights from Autoresearch for Solar Panel Segmentation", "published": "2026-10-07T17:45:31Z", "decision": "reviewed_deferred", "reason": "autoresearch编码agent仅改trainingprogram，评测split冻结，一小时GPU×3campaign；太阳板val改善不跨硬件，真operator执行待集成。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "2 Method", "location": ".cache/oct12-scan/fulltext/2610.10491.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "3 Experimental study", "location": ".cache/oct12-scan/fulltext/2610.10491.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 174, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10491v1", "sha256": "c24d008479e65bea92d0ff18b75393781e018ad8bba854dfebcd969771e33ba7", "local_path": ".cache/oct12-scan/fulltext/2610.10491.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10491v1", "result": "retrieved", "sha256": "c24d008479e65bea92d0ff18b75393781e018ad8bba854dfebcd969771e33ba7" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10491", "sha256": "a4e9c30de502f58a17103803d16ff94d984dfedcb7557a85be9535e74bb0ab3e", "local_path": ".cache/oct12-scan/fulltext/2610.10491.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 17:45:31 UTC (34 KB)", "published": "2026-10-07T17:45:31Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10478", "title": "Before They Can Solve: Predicting Post-Training Coding-Agent Performance from Base Models", "published": "2026-10-07T17:37:19Z", "reviews": [ { "id": "2610.10478", "title": "Before They Can Solve: Predicting Post-Training Coding-Agent Performance from Base Models", "published": "2026-10-07T17:37:19Z", "decision": "reviewed_deferred", "reason": "已读base-model screen方法/实验：成功trajectory逐patch真实test发现decisive step，BPB/MCQ/prefixpassK预测posttrained排名。特权prefix是明示checkpoint诊断，不是base冷启动agent能力，下一批严分screen相关性与真实posttraining作用。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Trajectory-Derived Protocols", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10478v1", "sha256": "3a4b0cf495e2d5ff45a5c692c18559b0686079cf847b118ba93f931c22fc660b", "local_html_sha256": "3a4b0cf495e2d5ff45a5c692c18559b0686079cf847b118ba93f931c22fc660b", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10478v1", "result": "retrieved", "sha256": "3a4b0cf495e2d5ff45a5c692c18559b0686079cf847b118ba93f931c22fc660b" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10448", "title": "MemoCare: An Interactive Multimodal Mobile System for Automated Cognitive Screening", "published": "2026-10-07T17:19:46Z", "reviews": [ { "id": "2610.10448", "title": "MemoCare: An Interactive Multimodal Mobile System for Automated Cognitive Screening", "published": "2026-10-07T17:19:46Z", "decision": "rejected_or_out_of_scope", "reason": "MemoCare移动确定性speech/GPS/touch评分与专用CNNscreening；151unitcase/71drawing并非LLM能力或临床有效性。", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "read_scope": "abstract_scope_exclusion; original HTML transport is not counted as reading", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 175, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10448v1", "sha256": "75a9e75868fca765dbfd7216361bbefa5a62724141b9cc0476c2aeccedecc5fd", "local_path": ".cache/oct12-scan/fulltext/2610.10448.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10448v1", "result": "retrieved", "sha256": "75a9e75868fca765dbfd7216361bbefa5a62724141b9cc0476c2aeccedecc5fd" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10448", "sha256": "9623d7435d0b422ab5d6f0a8c352aee0dda53fa192c62d383be018cd74e9eba8", "local_path": ".cache/oct12-scan/fulltext/2610.10448.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 17:19:46 UTC (255 KB)", "published": "2026-10-07T17:19:46Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10444", "title": "RunningTab: Direct Workspace Interaction with Environment-Side Tabs", "published": "2026-10-07T17:16:56Z", "reviews": [ { "id": "2610.10444", "title": "RunningTab: Direct Workspace Interaction with Environment-Side Tabs", "published": "2026-10-07T17:16:56Z", "decision": "reviewed_deferred", "reason": "已读RunningTab方法/实验：environment记录读过excerpt provenance与未打开candidate，agent维护requirements并finish check；三真实workspacebench/三models。下一批需真实file read事件与交付评分，不能用goldrequirement/答案预装tab。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10444v1", "sha256": "d66855986da2d75ad10552d4a7929cdf2850e62385da22cda35b196db5c78477", "local_html_sha256": "d66855986da2d75ad10552d4a7929cdf2850e62385da22cda35b196db5c78477", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10444v1", "result": "retrieved", "sha256": "d66855986da2d75ad10552d4a7929cdf2850e62385da22cda35b196db5c78477" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10436", "title": "Detecting Adversarial Images through Response Profiles of Vision-Language Models", "published": "2026-10-07T17:13:12Z", "reviews": [ { "id": "2610.10436", "title": "Detecting Adversarial Images through Response Profiles of Vision-Language Models", "published": "2026-10-07T17:13:12Z", "decision": "reviewed_deferred", "reason": "frozenVLM semanticprompt responseprofiles统计/图/cleanresidual/transformstability detector，跨attack/adaptive评测；需真实adv样本和trainonly统计。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "III Method", "location": ".cache/oct12-scan/fulltext/2610.10436.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "IV Experimental setup", "location": ".cache/oct12-scan/fulltext/2610.10436.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 176, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10436v1", "sha256": "d10e5172e5bc0f5b4612c34ed0f4797613a9e6e10f20be3a4e07fc336b456edf", "local_path": ".cache/oct12-scan/fulltext/2610.10436.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10436v1", "result": "retrieved", "sha256": "d10e5172e5bc0f5b4612c34ed0f4797613a9e6e10f20be3a4e07fc336b456edf" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10436", "sha256": "8f57d6fde69ff75d03a2a4ffa7197660df04f2c62e7484e427ee527393afb240", "local_path": ".cache/oct12-scan/fulltext/2610.10436.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 17:13:12 UTC (2,762 KB)", "published": "2026-10-07T17:13:12Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10426", "title": "CoTrace: Data Recipes for Training Terminal Agents with Harness-Model Co-Evolution", "published": "2026-10-07T17:06:59Z", "reviews": [ { "id": "2610.10426", "title": "CoTrace: Data Recipes for Training Terminal Agents with Harness-Model Co-Evolution", "published": "2026-10-07T17:06:59Z", "decision": "reviewed_deferred", "reason": "CoTrace componentwisepromotion的harness/model交替loop，以runtimeprovenance匹配verifiedSFT或freshRL，Tmax promotion/evolve/reservoir互斥与OODtransfer；下一批需真实terminalcontainers、控制器execute/evaluate闭环。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "2 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10426v1", "sha256": "8fd0c440501cf7ec0672b97cb4d8b173566b3b5dfd3e76a6fe76d0126d2c99d5", "local_html_sha256": "8fd0c440501cf7ec0672b97cb4d8b173566b3b5dfd3e76a6fe76d0126d2c99d5", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10426v1", "result": "retrieved", "sha256": "8fd0c440501cf7ec0672b97cb4d8b173566b3b5dfd3e76a6fe76d0126d2c99d5" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10659", "title": "Applying Security by Design at the Point of Execution: How Governed Security Requirements Affect the Security of AI-Generated Code", "published": "2026-10-07T17:02:09Z", "reviews": [ { "id": "2610.10659", "title": "Applying Security by Design at the Point of Execution: How Governed Security Requirements Affect the Security of AI-Generated Code", "published": "2026-10-07T17:02:09Z", "decision": "reviewed_deferred", "reason": "补读flattened原文4000-6500/10000-12500实验协议：治理security requirements由MCP交付无taskcode，DualGaugejudge与BaxBench真容器exploit比较；单generation、joint secure-functional无显著改善、manualKB未经专家审。下一批保留协议差异和负结果，不把oracle reminder matched效果当安全保证。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "Applying Security by Design at the Point of Execution: How\nGoverned Security Requirements Affect the Security of AI-Generated\nCode", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "Applying Security by Design at the Point of Execution: How\nGoverned Security Requirements Affect the Security of AI-Generated\nCode", "scope": "bounded flattened article hypothesis/experiment excerpts", "characters_read_at_least": [ 4000, 6500 ] }, { "heading": "Applying Security by Design at the Point of Execution: How\nGoverned Security Requirements Affect the Security of AI-Generated\nCode", "scope": "bounded flattened article protocol excerpt", "characters_read_at_least": [ 10000, 12500 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10659v1", "sha256": "6cdd88b9bebe9992831ea96fc068ce758a12149841345435c7e1e9fd45288eba", "local_html_sha256": "6cdd88b9bebe9992831ea96fc068ce758a12149841345435c7e1e9fd45288eba", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10659v1", "result": "retrieved", "sha256": "6cdd88b9bebe9992831ea96fc068ce758a12149841345435c7e1e9fd45288eba" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10422", "title": "Which Rollout Taught It That? BehaviorTrace and the Limits of Training-Data Attribution in Online RL", "published": "2026-10-07T17:01:37Z", "reviews": [ { "id": "2610.10422", "title": "Which Rollout Taught It That? BehaviorTrace and the Limits of Training-Data Attribution in Online RL", "published": "2026-10-07T17:01:37Z", "decision": "reviewed_deferred", "reason": "BehaviorTrace GRPO plantedtrigger causalgroundtruth+fullgradient sketching对比norm/fluency/headroom/seedcontrols；作者不提新attribution estimator，下一批diagnostic harness不得把plantedlabel提供policy或当能力提升。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Methods", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Results", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10422v1", "sha256": "468af7a5c0251545333a035b907a9718b8070849b49c1347dc90a2bca220488d", "local_html_sha256": "468af7a5c0251545333a035b907a9718b8070849b49c1347dc90a2bca220488d", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10422v1", "result": "retrieved", "sha256": "468af7a5c0251545333a035b907a9718b8070849b49c1347dc90a2bca220488d" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10411", "title": "Training Parallel Speculative Draft Models by Directly Minimizing Expected Decoding Rounds", "published": "2026-10-07T16:56:29Z", "reviews": [ { "id": "2610.10411", "title": "Training Parallel Speculative Draft Models by Directly Minimizing Expected Decoding Rounds", "published": "2026-10-07T16:56:29Z", "decision": "reviewed_deferred", "reason": "EDR用speculativedecoding Markovrewardprocess stateoccupancy精确expectedrounds，TDgradient训parallelDSpark/DFlydraft；九benchmark下一批需真实rollout/drafter与decoderoundevaluator，理论MainResults不代替实测。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Main Results", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10411v1", "sha256": "38bb29d63922e2068c4a8561287cbc2603d2acf7cd85ab281bc91c8272c53174", "local_html_sha256": "38bb29d63922e2068c4a8561287cbc2603d2acf7cd85ab281bc91c8272c53174", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10411v1", "result": "retrieved", "sha256": "38bb29d63922e2068c4a8561287cbc2603d2acf7cd85ab281bc91c8272c53174" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10407", "title": "SOTA: Stock Options Trading Agents Guided by Option-Implied Return Distributions", "published": "2026-10-07T16:55:05Z", "reviews": [ { "id": "2610.10407", "title": "SOTA: Stock Options Trading Agents Guided by Option-Implied Return Distributions", "published": "2026-10-07T16:55:05Z", "decision": "reviewed_deferred", "reason": "SOTA SFT→RL训练strategy-level期权选择，确定resolver执行合约/仓位，2024训练与2025 out-of-sample时间隔离明确；交易数据部分为收费源，财务结果不可泛化为投资建议，冻结批次外下一批先解决数据可得性。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 SOTA: Structured Strategy Selection for Stock Option Trading Agents", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10407v1", "sha256": "2053c3a903955b1962456713e90cc5b9d270b92f0b70a2d70684b11e38251fe2", "local_html_sha256": "2053c3a903955b1962456713e90cc5b9d270b92f0b70a2d70684b11e38251fe2", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10407v1", "result": "retrieved", "sha256": "2053c3a903955b1962456713e90cc5b9d270b92f0b70a2d70684b11e38251fe2" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10400", "title": "Self-correction Optimization for Interleaved Multimodal Generation", "published": "2026-10-07T16:51:02Z", "reviews": [ { "id": "2610.10400", "title": "Self-correction Optimization for Interleaved Multimodal Generation", "published": "2026-10-07T16:51:02Z", "decision": "reviewed_deferred", "reason": "SCO对CFG参考最小校正new-event/statepreservation，OpenING/ISG/VideoCraft；真实interleavedlatent denoise与成本待核。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Methodology", "location": ".cache/oct12-scan/fulltext/2610.10400.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.10400.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 177, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10400v1", "sha256": "02a6403909d2ccb14764e2a66c19c2e275b3f6b44acc0a6aef36864c7b25ba14", "local_path": ".cache/oct12-scan/fulltext/2610.10400.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10400v1", "result": "retrieved", "sha256": "02a6403909d2ccb14764e2a66c19c2e275b3f6b44acc0a6aef36864c7b25ba14" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10400", "sha256": "4997e651876a67dffd4fcc039bb79d07d37c057cc35a56b4ebf8629eb82e7f37", "local_path": ".cache/oct12-scan/fulltext/2610.10400.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 16:51:02 UTC (6,089 KB)", "published": "2026-10-07T16:51:02Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10395", "title": "Executing Causal Structure Learning with Linear-Attention Transformers", "published": "2026-10-07T16:48:51Z", "reviews": [ { "id": "2610.10395", "title": "Executing Causal Structure Learning with Linear-Attention Transformers", "published": "2026-10-07T16:48:51Z", "decision": "reviewed_deferred", "reason": "固定权linearattentionblock精确执行causaloptimizer，保留graph/multiplier；回放继承solver成败，是机制diagnostic而非学会causaldiscovery。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "8 Experiments", "location": ".cache/oct12-scan/fulltext/2610.10395.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "4 How the Transformer Executes an Update", "section_index": 3, "characters_read": [ 0, 1600 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 178, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10395v1", "sha256": "3554f59918c73c582639078ec162322c2c52ae8fad64e7e39ba14e6dedd4019e", "local_path": ".cache/oct12-scan/fulltext/2610.10395.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10395v1", "result": "retrieved", "sha256": "3554f59918c73c582639078ec162322c2c52ae8fad64e7e39ba14e6dedd4019e" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10395", "sha256": "746eca80105ec4233ceaf8321a6c4b44eabb8463061cc3341c476f09173c5c49", "local_path": ".cache/oct12-scan/fulltext/2610.10395.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 16:48:51 UTC (390 KB)", "published": "2026-10-07T16:48:51Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10390", "title": "Explicit Geometric Chain-of-Thought for Vision-Language-Action in Autonomous Driving", "published": "2026-10-07T16:47:17Z", "reviews": [ { "id": "2610.10390", "title": "Explicit Geometric Chain-of-Thought for Vision-Language-Action in Autonomous Driving", "published": "2026-10-07T16:47:17Z", "decision": "reviewed_deferred", "reason": "GeoCoTDrive先2Ddecisionregion再localized3Dpriors入ARtrajectory；nuScenesopenloop和NAVSIM/Bench2Driveclosedloop分明，非productionAB。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Method", "location": ".cache/oct12-scan/fulltext/2610.10390.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5 Experiments", "location": ".cache/oct12-scan/fulltext/2610.10390.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 179, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10390v1", "sha256": "4b36939ae0430956d3f4bdac263854e80cd96ba9c2a90c535a85c466a1997362", "local_path": ".cache/oct12-scan/fulltext/2610.10390.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10390v1", "result": "retrieved", "sha256": "4b36939ae0430956d3f4bdac263854e80cd96ba9c2a90c535a85c466a1997362" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10390", "sha256": "13ce8e3431cfd319d580c3e1c14597ce4386ca6b3ad3aa4736973afbadc1d06d", "local_path": ".cache/oct12-scan/fulltext/2610.10390.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 16:47:17 UTC (20,536 KB)", "published": "2026-10-07T16:47:17Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10385", "title": "OrBIT: Structure-Guided Embedding Compression", "published": "2026-10-07T16:44:47Z", "reviews": [ { "id": "2610.10385", "title": "OrBIT: Structure-Guided Embedding Compression", "published": "2026-10-07T16:44:47Z", "decision": "reviewed_deferred", "reason": "OrBIT orbit共享codeword/gluing/globalresidual预算，GPT2与三7Bembedding rate-distortion；需补end-to-endLM和真实codec成本。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 ORBIT-STRUCTURED ADDITIVE COMPRESSION AND FRAME GLUING", "location": ".cache/oct12-scan/fulltext/2610.10385.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 EXPERIMENTS", "location": ".cache/oct12-scan/fulltext/2610.10385.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 180, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10385v1", "sha256": "255071e3b046f586d6d47f11983511cd0e89e586883bfab708fd1f6a29115376", "local_path": ".cache/oct12-scan/fulltext/2610.10385.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10385v1", "result": "retrieved", "sha256": "255071e3b046f586d6d47f11983511cd0e89e586883bfab708fd1f6a29115376" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10385", "sha256": "6e37710b9d5f2d4f8a2ff97fa1af32153893f6848ff04e31132c81374f4bab05", "local_path": ".cache/oct12-scan/fulltext/2610.10385.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 16:44:47 UTC (124 KB)", "published": "2026-10-07T16:44:47Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10378", "title": "Document-Level Text Simplification in Estonian Using Large Language Models", "published": "2026-10-07T16:38:26Z", "reviews": [ { "id": "2610.10378", "title": "Document-Level Text Simplification in Estonian Using Large Language Models", "published": "2026-10-07T16:38:26Z", "decision": "reviewed_deferred", "reason": "已读Estonian simplification prompt策略/实验：singlepass/modularagent/guideline workflow，15短Wikipedia文章及独立linguistreference/manualcoherence。小样本语言应用保留，非长程工具agent，暂不优先实现/不泛化到全document能力。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Large Language Model Prompting Strategies for Document Simplification", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10378v1", "sha256": "7b05c3b8b7cf5bc6f4315714743014f16c71dc52b72548d7404d8b20707f1be0", "local_html_sha256": "7b05c3b8b7cf5bc6f4315714743014f16c71dc52b72548d7404d8b20707f1be0", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10378v1", "result": "retrieved", "sha256": "7b05c3b8b7cf5bc6f4315714743014f16c71dc52b72548d7404d8b20707f1be0" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10655", "title": "Nullify: Null-Space Activation Steering for Training-Free LLM Unlearning", "published": "2026-10-07T16:24:07Z", "reviews": [ { "id": "2610.10655", "title": "Nullify: Null-Space Activation Steering for Training-Free LLM Unlearning", "published": "2026-10-07T16:24:07Z", "decision": "reviewed_deferred", "reason": "Nullify forgetsteering约束retainnullspace，TOFU/MUSE；可拔除activationintervention不等于权重事实擦除，keepcalibration与test需隔离。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Methodology", "location": ".cache/oct12-scan/fulltext/2610.10655.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.10655.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 181, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10655v1", "sha256": "5df0f79531df4b735a9ed8c2bb5adaf7315482505fba5fbd957ef9fac6a3a64a", "local_path": ".cache/oct12-scan/fulltext/2610.10655.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10655v1", "result": "retrieved", "sha256": "5df0f79531df4b735a9ed8c2bb5adaf7315482505fba5fbd957ef9fac6a3a64a" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10655", "sha256": "892878eda6587cf04dbf12fbbab919c2c6feb31795a87abb51b818ff9dd61b2e", "local_path": ".cache/oct12-scan/fulltext/2610.10655.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 16:24:07 UTC (80 KB)", "published": "2026-10-07T16:24:07Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10323", "title": "HAN-Mamba: Hierarchical Selective State Space Networks for Multi-Scale Financial Volatility Forecasting", "published": "2026-10-07T16:12:52Z", "reviews": [ { "id": "2610.10323", "title": "HAN-Mamba: Hierarchical Selective State Space Networks for Multi-Scale Financial Volatility Forecasting", "published": "2026-10-07T16:12:52Z", "decision": "rejected_or_out_of_scope", "reason": "HAN-Mamba多频率金融volatility专用Optiver预测架构，非本批LLM/foundation后训练方法。", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "read_scope": "abstract_scope_exclusion; original HTML transport is not counted as reading", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 182, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10323v1", "sha256": "50542ab4139e283f5979d43e3d5aa1ba8cb4369d602547f98200a38bd18dc908", "local_path": ".cache/oct12-scan/fulltext/2610.10323.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10323v1", "result": "retrieved", "sha256": "50542ab4139e283f5979d43e3d5aa1ba8cb4369d602547f98200a38bd18dc908" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10323", "sha256": "60965c761f5512e5b4d759aa3c93ce1f455b62cdd8de24b668f8e38ff43be9c2", "local_path": ".cache/oct12-scan/fulltext/2610.10323.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 16:12:52 UTC (28 KB)", "published": "2026-10-07T16:12:52Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10320", "title": "LLM-Assisted Generation of Transparent, Open-Source Multiphysics Models of Electrochemical Devices", "published": "2026-10-07T16:11:56Z", "reviews": [ { "id": "2610.10320", "title": "LLM-Assisted Generation of Transparent, Open-Source Multiphysics Models of Electrochemical Devices", "published": "2026-10-07T16:11:56Z", "decision": "reviewed_deferred", "reason": "已读electrochemical Julia model方法/结果：人写物理equations/参数/数值阶段/可审checkpoint，agent真实build与COMSOL及independentbuild对照，植错debug。下一批科学codeeval可行但gold numerical解只外部评分，不以专用spec当自主发现物理能力。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "Methods", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10320v1", "sha256": "98585fa11c69a9f595d53a0b2ddd2ff27c83e649a5d9333b6dec2762f8b23bbd", "local_html_sha256": "98585fa11c69a9f595d53a0b2ddd2ff27c83e649a5d9333b6dec2762f8b23bbd", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10320v1", "result": "retrieved", "sha256": "98585fa11c69a9f595d53a0b2ddd2ff27c83e649a5d9333b6dec2762f8b23bbd" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10311", "title": "Fault-tolerant foundation models", "published": "2026-10-07T16:05:48Z", "reviews": [ { "id": "2610.10311", "title": "Fault-tolerant foundation models", "published": "2026-10-07T16:05:48Z", "decision": "reviewed_deferred", "reason": "faultinjection在matmuloperand按block随机zero训练LLM scaling，FineWebheldout与多seed；40kGPUhour模拟故障，节能/商业尺度为推测。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "Methods", "location": ".cache/oct12-scan/fulltext/2610.10311.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "Discussion", "section_index": 0, "characters_read": [ 0, 1600 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "Methods", "section_index": 1, "characters_read": [ 0, 6500 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 183, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10311v1", "sha256": "8f140805051908a5bbe4fa1bd3c77241387ab40c02ff65bf7a79f3afca5e5005", "local_path": ".cache/oct12-scan/fulltext/2610.10311.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10311v1", "result": "retrieved", "sha256": "8f140805051908a5bbe4fa1bd3c77241387ab40c02ff65bf7a79f3afca5e5005" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10311", "sha256": "2cc546f3f1d541430acfc32a70ca8eff3d647dd009d5c6d38c29939d3f61e5fa", "local_path": ".cache/oct12-scan/fulltext/2610.10311.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 16:05:48 UTC (6,366 KB)", "published": "2026-10-07T16:05:48Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10305", "title": "How Do Transformers Learn to Represent Symmetries?", "published": "2026-10-07T16:01:51Z", "reviews": [ { "id": "2610.10305", "title": "How Do Transformers Learn to Represent Symmetries?", "published": "2026-10-07T16:01:51Z", "decision": "reviewed_deferred", "reason": "Transformerfiniteaugmentation学习pointcloud群不变性与equivariance，受控architecture/ablation；真实训练/OOD几何证据非fixture待实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "additional targeted method/evaluation", "heading": "3 Finite augmentation reveals a symmetry inductive bias in Transformers", "section_index": 2, "characters_read": [ 0, 1602 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "5 Mechanisms that can induce invariance to angle-preserving symmetries", "section_index": 4, "characters_read": [ 0, 1625 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 184, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10305v1", "sha256": "24d5bba69a473a81d6578d932193b54d6ba1a45e98e444c422fa4c9333ff410a", "local_path": ".cache/oct12-scan/fulltext/2610.10305.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10305v1", "result": "retrieved", "sha256": "24d5bba69a473a81d6578d932193b54d6ba1a45e98e444c422fa4c9333ff410a" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10305", "sha256": "5383016a4c19e7b34cf1fbe15eb2156fa9c3d9cf806627b7f20fcb81ec4aad35", "local_path": ".cache/oct12-scan/fulltext/2610.10305.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 16:01:51 UTC (4,542 KB)", "published": "2026-10-07T16:01:51Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10304", "title": "SemanticFold: Latent Sequence Compression SeparatesLanguage Modeling, Decodability, and Reasoning", "published": "2026-10-07T16:00:59Z", "reviews": [ { "id": "2610.10304", "title": "SemanticFold: Latent Sequence Compression SeparatesLanguage Modeling, Decodability, and Reasoning", "published": "2026-10-07T16:00:59Z", "decision": "reviewed_deferred", "reason": "SemanticFold prefixlatentfold/learnedboundary，fixedtargetcontinuation排除target选择；NLL、probe、reasoning与系统端点不一致，未本地测量。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "4 Experimental Design", "location": ".cache/oct12-scan/fulltext/2610.10304.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experimental Design", "location": ".cache/oct12-scan/fulltext/2610.10304.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 SemanticFold", "section_index": 2, "characters_read": [ 0, 1800 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 185, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10304v1", "sha256": "275fdf3610e55ada6666bdce226ac1a7ca47c8748b0e4ca4353be58169c5598b", "local_path": ".cache/oct12-scan/fulltext/2610.10304.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10304v1", "result": "retrieved", "sha256": "275fdf3610e55ada6666bdce226ac1a7ca47c8748b0e4ca4353be58169c5598b" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10304", "sha256": "d8266e9b3e9df2e39c57309632e3bd1e8c1981807f4b7950f32663100079716c", "local_path": ".cache/oct12-scan/fulltext/2610.10304.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 16:00:59 UTC (32 KB)", "published": "2026-10-07T16:00:59Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10302", "title": "Continual Graph Multi-Agent Reinforcement Learning", "published": "2026-10-07T15:56:31Z", "reviews": [ { "id": "2610.10302", "title": "Continual Graph Multi-Agent Reinforcement Learning", "published": "2026-10-07T15:56:31Z", "decision": "rejected_or_out_of_scope", "reason": "已读CGMARL/GRAFO benchmark与prelim：taskgraph决定传统cooperative导航dynamics/人数，frozen graphencoder配continuallearning防forgetting；不是LLM agent，本批范围外。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 GRAFO Benchmark", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 Preliminaries", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10302v1", "sha256": "7ac0ca2b1b3e1f736c3636231fec1fe170f2347ff1c42ada3321cff3b4f10dc2", "local_html_sha256": "7ac0ca2b1b3e1f736c3636231fec1fe170f2347ff1c42ada3321cff3b4f10dc2", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10302v1", "result": "retrieved", "sha256": "7ac0ca2b1b3e1f736c3636231fec1fe170f2347ff1c42ada3321cff3b4f10dc2" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10286", "title": "$Δ$Representation: Geometry Supervised Representation Learning of Phenotypes via Counterfactual Reasoning for Medical VLMs", "published": "2026-10-07T15:49:36Z", "reviews": [ { "id": "2610.10286", "title": "$Δ$Representation: Geometry Supervised Representation Learning of Phenotypes via Counterfactual Reasoning for Medical VLMs", "published": "2026-10-07T15:49:36Z", "decision": "reviewed_deferred", "reason": "DeltaRepresentation解剖几何+lesion/normalcounterfactualincrement聚类，ReXCT/LIDC；真实medicalVLM与patientdisjoint/标签访问待核。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Methodology", "location": ".cache/oct12-scan/fulltext/2610.10286.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.10286.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 186, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10286v1", "sha256": "89ffdb2550d76c121cf8ac45ee285d94dfdc982f0e1da7ed9dab3baed5c1f9c7", "local_path": ".cache/oct12-scan/fulltext/2610.10286.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10286v1", "result": "retrieved", "sha256": "89ffdb2550d76c121cf8ac45ee285d94dfdc982f0e1da7ed9dab3baed5c1f9c7" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10286", "sha256": "69d37d38936262de714f564d6906ffed3bd0bdd94d31e62b2da5fe55b0d87892", "local_path": ".cache/oct12-scan/fulltext/2610.10286.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 15:49:36 UTC (3,111 KB)", "published": "2026-10-07T15:49:36Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10265", "title": "Stale, Misattributed, or Late: Where Personal Memory Fails Before Generation", "published": "2026-10-07T15:37:12Z", "reviews": [ { "id": "2610.10265", "title": "Stale, Misattributed, or Late: Where Personal Memory Fails Before Generation", "published": "2026-10-07T15:37:12Z", "decision": "reviewed_deferred", "reason": "已读personal memory摘要/experimental setup及LLMuse：PFM分state temporalvalidity、keymerge/identity、servinglatency；controlledrevisions+LongMemEval/LoCoMo，开放merge recall极低。下一批需源slot赋值真实模型、不同人同名/旧值exposure评测，不能以answeraccuracy遮掩存储错误。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "Use of Large Language Models", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "6 Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10265v1", "sha256": "911a61a7cda3558ba7b229f5981edd32b2ce853bba1e155af584e13c7ebac995", "local_html_sha256": "911a61a7cda3558ba7b229f5981edd32b2ce853bba1e155af584e13c7ebac995", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10265v1", "result": "retrieved", "sha256": "911a61a7cda3558ba7b229f5981edd32b2ce853bba1e155af584e13c7ebac995" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10263", "title": "When Sub-Agents Work in Parallel: The Promises and Pitfalls of Dynamic Concurrency in Long-Horizon Coding Tasks", "published": "2026-10-07T15:35:22Z", "reviews": [ { "id": "2610.10263", "title": "When Sub-Agents Work in Parallel: The Promises and Pitfalls of Dynamic Concurrency in Long-Horizon Coding Tasks", "published": "2026-10-07T15:35:22Z", "decision": "reviewed_deferred", "reason": "已读dynamic concurrency方法/效用结果：354task2124 matched Codex/Claude/Kimi enabled/disabled，13failuremode、end-to-end成本/quality。下一批evalinfra要权限限定并记录subagent协调/重复与失败成本，不据论文另行授权spawn。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3. Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4. Effectiveness and Efficiency (RQ1)", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10263v1", "sha256": "e5cb326ee79d5bb4d26dfa109bc70537d80dce3b2ed06e5f1efdd98b4ceac10b", "local_html_sha256": "e5cb326ee79d5bb4d26dfa109bc70537d80dce3b2ed06e5f1efdd98b4ceac10b", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10263v1", "result": "retrieved", "sha256": "e5cb326ee79d5bb4d26dfa109bc70537d80dce3b2ed06e5f1efdd98b4ceac10b" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10258", "title": "QuSema: Detecting Silent Bugs in Quantum Libraries via Quantum-knowledge-enhanced Agents", "published": "2026-10-07T15:33:38Z", "reviews": [ { "id": "2610.10258", "title": "QuSema: Detecting Silent Bugs in Quantum Libraries via Quantum-knowledge-enhanced Agents", "published": "2026-10-07T15:33:38Z", "decision": "reviewed_deferred", "reason": "已读QuSema quantum semantic evidence与experimental setup：源码/文档语义hypothesis经真实Qiskit/PennyLane API test确认，20历史bug+40authorconfirmednewbug。下一批需真实库版本与可触发测试，不把LM semantic判断当oracle正确性。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "5. Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3. Current Limitations and Semantic Evidence for Quantum Silent Bug Detection", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10258v1", "sha256": "2add8bbc19dd2ca6e92dc487652f967b3eedac36c01df1800d9a7411ca077a1d", "local_html_sha256": "2add8bbc19dd2ca6e92dc487652f967b3eedac36c01df1800d9a7411ca077a1d", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10258v1", "result": "retrieved", "sha256": "2add8bbc19dd2ca6e92dc487652f967b3eedac36c01df1800d9a7411ca077a1d" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10256", "title": "OOM-RL II: Reality Is an Oracle, Not a Debugger Provenance-Constrained Diagnosis in Continually Evolving Agent-Engineered Systems", "published": "2026-10-07T15:33:17Z", "reviews": [ { "id": "2610.10256", "title": "OOM-RL II: Reality Is an Oracle, Not a Debugger Provenance-Constrained Diagnosis in Continually Evolving Agent-Engineered Systems", "published": "2026-10-07T15:33:17Z", "decision": "reviewed_deferred", "reason": "已读OOM-RL II provenance framework/evaluation：一真实tradingaccount过程有收益但annual alpha不显著，recommendation-runtime绑定不全、无共同冻结实例；prospective protocol只提议。保留研究integrity参考，不用结果归因agent性能或投资建议。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 Conceptual Framework: Reality as Oracle, Not Debugger", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "7 RQ3: Provenance-Bounded Evaluation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10256v1", "sha256": "380476e65a03be5ad3a4c1d5fe377ec79dd5c31bcc02a8bb44c5dace76248ddb", "local_html_sha256": "380476e65a03be5ad3a4c1d5fe377ec79dd5c31bcc02a8bb44c5dace76248ddb", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10256v1", "result": "retrieved", "sha256": "380476e65a03be5ad3a4c1d5fe377ec79dd5c31bcc02a8bb44c5dace76248ddb" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10242", "title": "TestGRAD: Evolving Test Suites via Failure Pattern Momentum for SWE-Agent Ensemble", "published": "2026-10-07T15:25:15Z", "reviews": [ { "id": "2610.10242", "title": "TestGRAD: Evolving Test Suites via Failure Pattern Momentum for SWE-Agent Ensemble", "published": "2026-10-07T15:25:15Z", "decision": "reviewed_deferred", "reason": "已读TestGRAD framework/setting：真实候选patch differentialtest loss、CRUD edits与failure-patternmemory，LLM gradient/momentum是执行反馈类比非微分训练；6leaderboardpatchpool/SWE500外部分数。下一批不得用goldresolved标签选择patch/删正确约束。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "TestGRAD Framework", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10242v1", "sha256": "b93c0c4db255feec8c72be024559758b6fae5ebad8933637a32b8a83e35ece60", "local_html_sha256": "b93c0c4db255feec8c72be024559758b6fae5ebad8933637a32b8a83e35ece60", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10242v1", "result": "retrieved", "sha256": "b93c0c4db255feec8c72be024559758b6fae5ebad8933637a32b8a83e35ece60" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10238", "title": "Geometry-Supervised Visual Representation Learning for Multi-Phenotype Lesion Interpretation in Medical VLMs", "published": "2026-10-07T15:22:11Z", "reviews": [ { "id": "2610.10238", "title": "Geometry-Supervised Visual Representation Learning for Multi-Phenotype Lesion Interpretation in Medical VLMs", "published": "2026-10-07T15:22:11Z", "decision": "reviewed_deferred", "reason": "PureEyes理想解剖/phenotypegeometry+PureNeurons sparseevidence软token，CT/mammogram/MRI公开bench；实际image模型未实施。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Methodology", "location": ".cache/oct12-scan/fulltext/2610.10238.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.10238.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 187, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10238v1", "sha256": "dbc3519dc4f1c37345d28705410088f77d227ff46dbb345aac0efa1cbbcb916a", "local_path": ".cache/oct12-scan/fulltext/2610.10238.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10238v1", "result": "retrieved", "sha256": "dbc3519dc4f1c37345d28705410088f77d227ff46dbb345aac0efa1cbbcb916a" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10238", "sha256": "75d5cd8b6f6ed12ce371cb44144a9f20c782aadc9e7f7b78258967a3c8872162", "local_path": ".cache/oct12-scan/fulltext/2610.10238.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 15:22:11 UTC (6,034 KB)", "published": "2026-10-07T15:22:11Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10226", "title": "Why Software Engineering Is Indispensable in the Age of Coding Agents", "published": "2026-10-07T15:17:37Z", "reviews": [ { "id": "2610.10226", "title": "Why Software Engineering Is Indispensable in the Age of Coding Agents", "published": "2026-10-07T15:17:37Z", "decision": "rejected_or_out_of_scope", "reason": "已读SE indispensable论证段：persistent artifacts/human methodology的立场论文，无独立新执行算法/controlledbenchmark；不采纳其概括性结构结论作实证，范围外工程参考。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "1. What We See", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2. Why: The Structural Vacuum", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10226v1", "sha256": "6b653ba000c219f284ef445fbf643369402fcbc39324b9c46a6f9e76c2e1db5f", "local_html_sha256": "6b653ba000c219f284ef445fbf643369402fcbc39324b9c46a6f9e76c2e1db5f", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10226v1", "result": "retrieved", "sha256": "6b653ba000c219f284ef445fbf643369402fcbc39324b9c46a6f9e76c2e1db5f" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10206", "title": "Computations of the slice genus and the unknotting number of links via machine learning", "published": "2026-10-07T15:04:04Z", "reviews": [ { "id": "2610.10206", "title": "Computations of the slice genus and the unknotting number of links via machine learning", "published": "2026-10-07T15:04:04Z", "decision": "rejected_or_out_of_scope", "reason": "已读link topology environment/basics：RL与Bayesian优化数学link invariant上界/下界，非LLM工具agent或基础模型实现，本批范围外传统科学RL。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2. Basics: links, diagrams, the band game and RL", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3. The environment", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10206v1", "sha256": "6db2bb83996254e35288c655621764d0ddec1b73dc68f3543b9c180efda490e4", "local_html_sha256": "6db2bb83996254e35288c655621764d0ddec1b73dc68f3543b9c180efda490e4", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10206v1", "result": "retrieved", "sha256": "6db2bb83996254e35288c655621764d0ddec1b73dc68f3543b9c180efda490e4" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10201", "title": "GAGR-Lab: Evaluating Joint Spatial-Geometric and Analytic Function Reasoning", "published": "2026-10-07T15:01:54Z", "reviews": [ { "id": "2610.10201", "title": "GAGR-Lab: Evaluating Joint Spatial-Geometric and Analytic Function Reasoning", "published": "2026-10-07T15:01:54Z", "decision": "reviewed_deferred", "reason": "GAGR-Lab Rust权威trajectory执行，72game hostedmodel0hits、privilegedanalyticcontrol600成功；24cellcomparative尚prospective，保留diagnostic。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "4 Capability decomposition and difficulty design", "location": ".cache/oct12-scan/fulltext/2610.10201.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5 Auditable evaluation and experimental questions", "location": ".cache/oct12-scan/fulltext/2610.10201.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 Task and execution semantics", "section_index": 2, "characters_read": [ 0, 1401 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "6 Executed protocol", "section_index": 5, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "7 Results", "section_index": 6, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 188, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10201v1", "sha256": "a31062c6c4ce9405a353cbfe2806c57b40ecbbe54c3dc97e892850fce95c5a67", "local_path": ".cache/oct12-scan/fulltext/2610.10201.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10201v1", "result": "retrieved", "sha256": "a31062c6c4ce9405a353cbfe2806c57b40ecbbe54c3dc97e892850fce95c5a67" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10201", "sha256": "ea9598128a08440f3ff805d010d3cfeb66c4d6931169d669f9c300e6aa9fa747", "local_path": ".cache/oct12-scan/fulltext/2610.10201.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 15:01:54 UTC (75 KB)", "published": "2026-10-07T15:01:54Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10184", "title": "Agentic AI-Assisted Modeling for Production Scheduling: Assessment in Constraint Programming", "published": "2026-10-07T14:51:37Z", "reviews": [ { "id": "2610.10184", "title": "Agentic AI-Assisted Modeling for Production Scheduling: Assessment in Constraint Programming", "published": "2026-10-07T14:51:37Z", "decision": "reviewed_deferred", "reason": "已读CP agentworkflow/MCP方法及full-factorial evaluation：formulation/coding分工，solverdoc检索、六工业风格调度三LLM，脚本runtime success与数学正确性分开。下一批需真实solver/execution，无toolbaseline使workflow与MCP联合效应不能归单agent。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experimental evaluation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10184v1", "sha256": "e4a04385e7013875be8e2a21ef65c11f14cdea176d8d3aac08693d9f816cd060", "local_html_sha256": "e4a04385e7013875be8e2a21ef65c11f14cdea176d8d3aac08693d9f816cd060", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10184v1", "result": "retrieved", "sha256": "e4a04385e7013875be8e2a21ef65c11f14cdea176d8d3aac08693d9f816cd060" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10183", "title": "VideoEvolve: Co-Evolving Memory and Retrieval for Long Video Understanding", "published": "2026-10-07T14:51:23Z", "reviews": [ { "id": "2610.10183", "title": "VideoEvolve: Co-Evolving Memory and Retrieval for Long Video Understanding", "published": "2026-10-07T14:51:23Z", "decision": "reviewed_deferred", "reason": "已读VideoEvolve方法/训练设置：实际Memory/RetrievalEvolver alternatedRL，一方冻结、BEF/CEF用disjoint诊断video集无gradient；冷启动SFT与视频帧访问。下一批multimodal候选必须真实视觉memory+双policy训练，固定summary/questionmemory非机制复现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiment", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10183v1", "sha256": "7d3fc9322f0b52d2d43a6d0890915b39372e158ff39a371b83518b42d138f29d", "local_html_sha256": "7d3fc9322f0b52d2d43a6d0890915b39372e158ff39a371b83518b42d138f29d", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10183v1", "result": "retrieved", "sha256": "7d3fc9322f0b52d2d43a6d0890915b39372e158ff39a371b83518b42d138f29d" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10179", "title": "Beyond Outcome Rewards: Constructing and Assigning Retrieval Credit for Search Agents", "published": "2026-10-07T14:48:47Z", "reviews": [ { "id": "2610.10179", "title": "Beyond Outcome Rewards: Constructing and Assigning Retrieval Credit for Search Agents", "published": "2026-10-07T14:48:47Z", "decision": "reviewed_deferred", "reason": "Search-agent intermediate retrieval progress credit比较scalar/local归属，固定E5/KILT retriever、14k prompts、5 rollouts同预算GRPO控制；核心检索credit需真实query/evidence事件，下一批研究。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Assigning retrieval credit to policy updates", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experimental design and results", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10179v1", "sha256": "9a693b49ea1ec48fc6772e5e4e153fa034ec2f4d09f19d2adcc455026f378ad2", "local_html_sha256": "9a693b49ea1ec48fc6772e5e4e153fa034ec2f4d09f19d2adcc455026f378ad2", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10179v1", "result": "retrieved", "sha256": "9a693b49ea1ec48fc6772e5e4e153fa034ec2f4d09f19d2adcc455026f378ad2" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10163", "title": "Beyond Anonymous Captions: Grounding Character Identity in Video Captioning and Question Answering", "published": "2026-10-07T14:41:29Z", "reviews": [ { "id": "2610.10163", "title": "Beyond Anonymous Captions: Grounding Character Identity in Video Captioning and Question Answering", "published": "2026-10-07T14:41:29Z", "decision": "reviewed_deferred", "reason": "BAC referenceface/tracks与box+coords、多尺度QLoRAvideo；人工verifiedmovieQA公开资产，电影授权/人脸reference与split待核。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "2 Methods", "location": ".cache/oct12-scan/fulltext/2610.10163.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "3 Results", "location": ".cache/oct12-scan/fulltext/2610.10163.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 189, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10163v1", "sha256": "5c4b63dba163b6be102ede1289107f7e2a3decc916c3fc4488195296dcea4e9f", "local_path": ".cache/oct12-scan/fulltext/2610.10163.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10163v1", "result": "retrieved", "sha256": "5c4b63dba163b6be102ede1289107f7e2a3decc916c3fc4488195296dcea4e9f" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10163", "sha256": "7faa90f2abe425a7c2bad6335a3261c37eee0190dcbffeaa448d7fc5e522abce", "local_path": ".cache/oct12-scan/fulltext/2610.10163.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 14:41:29 UTC (3,876 KB)", "published": "2026-10-07T14:41:29Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10641", "title": "Coverage-Aware Reasoning with Medical Tokens for Diagnosis Prediction", "published": "2026-10-07T14:39:18Z", "reviews": [ { "id": "2610.10641", "title": "Coverage-Aware Reasoning with Medical Tokens for Diagnosis Prediction", "published": "2026-10-07T14:39:18Z", "decision": "reviewed_deferred", "reason": "CARing ontology-enriched residual-quantized medical SID tokens、alignment+reasoning training、coverage RL和multi-positive监督，MIMIC3/4多标签评测；健康数据受准入协议限制，下一批应先满足data access与多label覆盖contract。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10641v1", "sha256": "38d8000670fab99702044d4287c0368f9bf2c728765a2b91de3dad25ed79cd22", "local_html_sha256": "38d8000670fab99702044d4287c0368f9bf2c728765a2b91de3dad25ed79cd22", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10641v1", "result": "retrieved", "sha256": "38d8000670fab99702044d4287c0368f9bf2c728765a2b91de3dad25ed79cd22" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10156", "title": "HeiCo-FOCUS: A Clinically Grounded Dataset for Long-Context Video Understanding", "published": "2026-10-07T14:33:47Z", "reviews": [ { "id": "2610.10156", "title": "HeiCo-FOCUS: A Clinically Grounded Dataset for Long-Context Video Understanding", "published": "2026-10-07T14:33:47Z", "decision": "reviewed_deferred", "reason": "HeiCoFOCUS hoursurgery30kQA多track、39clinicalexperts；10frontier模型/validatedjudge/textonly，longvideo和患者权限未实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "5 Experiments and Results", "location": ".cache/oct12-scan/fulltext/2610.10156.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "4 visual question answering Generation Pipeline", "section_index": 3, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 190, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10156v1", "sha256": "83973638fa85ea0c02e170595797da69cfdc52ba3948d3658efe007b90b8b050", "local_path": ".cache/oct12-scan/fulltext/2610.10156.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10156v1", "result": "retrieved", "sha256": "83973638fa85ea0c02e170595797da69cfdc52ba3948d3658efe007b90b8b050" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10156", "sha256": "396561d90c1c3408d2f39da79c6986f0fb4da6743530e22809079667434ac3b0", "local_path": ".cache/oct12-scan/fulltext/2610.10156.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 14:33:47 UTC (1,455 KB)", "published": "2026-10-07T14:33:47Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10639", "title": "Visible Reasoning Is Not a Universal Optimizer: Persona- and Thinking-Dependent Effects in Analytics Code Generation", "published": "2026-10-07T14:27:58Z", "reviews": [ { "id": "2610.10639", "title": "Visible Reasoning Is Not a Universal Optimizer: Persona- and Thinking-Dependent Effects in Analytics Code Generation", "published": "2026-10-07T14:27:58Z", "decision": "reviewed_deferred", "reason": "Optum在BIRD-Persona上的matched SQL/pandas、persona、CoT和内部思考配置比较；§3预注册304问题、§4主family252tests无BH发现，结论不支持通用CoT收益。可纳入执行型评测参考，非新推荐模型、非这16项实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "bounded_methods_and_evaluation", "review_artifact": "research-audits/oct12-review-closure/recommendation-reviews.json", "evidence_locations": [ "§3.1–3.3", "§4.1–4.4", "§7 Limitations" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10639v1", "sha256": "38e39c02e1a5ffee3c67f1a49b4902c08917291425076a418daf97152e243f0f" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "recommendation" ], "status": "reviewed_deferred" }, { "id": "2610.10150", "title": "On the Reliability of LLM-Based Vulnerability Patching Benchmarks", "published": "2026-10-07T14:26:57Z", "reviews": [ { "id": "2610.10150", "title": "On the Reliability of LLM-Based Vulnerability Patching Benchmarks", "published": "2026-10-07T14:26:57Z", "decision": "reviewed_deferred", "reason": "已读vulnerability patch benchmark taxonomy/data方法：112bug84repo，可单独toggle permissions/timeouts/infra，PoC/regression/developerdesign tests区分。下一批executioninfra重要，不能删除infra失败或把PoC symptom通过声称上游修复质量。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Methodology and Dataset", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 A Taxonomy of Benchmarking Pitfalls", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10150v1", "sha256": "0eea100e71c605df950d3731267baba63b193ce360332da0285f65b750e4faed", "local_html_sha256": "0eea100e71c605df950d3731267baba63b193ce360332da0285f65b750e4faed", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10150v1", "result": "retrieved", "sha256": "0eea100e71c605df950d3731267baba63b193ce360332da0285f65b750e4faed" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10145", "title": "InterView-C: A Synchronized Multimodal Corpus of VR Avatar-Mediated Survey Interviews", "published": "2026-10-07T14:23:32Z", "reviews": [ { "id": "2610.10145", "title": "InterView-C: A Synchronized Multimodal Corpus of VR Avatar-Mediated Survey Interviews", "published": "2026-10-07T14:23:32Z", "decision": "rejected_or_out_of_scope", "reason": "InterView-C主要贡献是German VR interview语料、ASR/negation transfer诊断；读摘要及study procedure，没有当前批次LLM posttraining/agent核心算法，不因Meta Quest设备误判Meta机构论文。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Experiment Setup", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10145v1", "sha256": "3d28484991584187b3a00b3f409dd848592a0996e447225d5fa67e7931a35830", "local_html_sha256": "3d28484991584187b3a00b3f409dd848592a0996e447225d5fa67e7931a35830", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10145v1", "result": "retrieved", "sha256": "3d28484991584187b3a00b3f409dd848592a0996e447225d5fa67e7931a35830" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10141", "title": "AdaT$^2$: Adaptive Test Transformations for Black-Box Boundary Testing of Conversational Agents", "published": "2026-10-07T14:19:02Z", "reviews": [ { "id": "2610.10141", "title": "AdaT$^2$: Adaptive Test Transformations for Black-Box Boundary Testing of Conversational Agents", "published": "2026-10-07T14:19:02Z", "decision": "reviewed_deferred", "reason": "已读AdaT2方法/evaluation：exploratoryreply statement+test transformation对，novelty选statement、BayesUCB按适用/结果更新，boundary双侧/明确authority与domain推断分开。下一批需真agent执行和judge校准，tests数差异控制不能按fault数直接比较。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3. Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4. Evaluation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10141v1", "sha256": "a19458a9ed33e0d09a3bd84cf7ea240edac37484571906f232dcdc32bc210ebc", "local_html_sha256": "a19458a9ed33e0d09a3bd84cf7ea240edac37484571906f232dcdc32bc210ebc", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10141v1", "result": "retrieved", "sha256": "a19458a9ed33e0d09a3bd84cf7ea240edac37484571906f232dcdc32bc210ebc" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10126", "title": "Know the Shape, Find the Fault: Topology-Conditioned Diagnosis of Multi-Agent LLM Failures", "published": "2026-10-07T14:07:48Z", "reviews": [ { "id": "2610.10126", "title": "Know the Shape, Find the Fault: Topology-Conditioned Diagnosis of Multi-Agent LLM Failures", "published": "2026-10-07T14:07:48Z", "decision": "reviewed_deferred", "reason": "已读MAScope empirical/experiment：message证据恢复communicationtopology、另labeltrace学failureprior，TCJudge分类；fixedtopology一次extraction amortization cost为projected。下一批需train/test prior隔离、predictedtopology与goldtopology诊断分开，不把关联chi2当causal。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "5. Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3. Empirical Foundation: Topology × \\times Failure Correlation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10126v1", "sha256": "3602ed489776409bc05a295a4c4b5227a5bc0b5403b3054fc7fb81256ab0eef6", "local_html_sha256": "3602ed489776409bc05a295a4c4b5227a5bc0b5403b3054fc7fb81256ab0eef6", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10126v1", "result": "retrieved", "sha256": "3602ed489776409bc05a295a4c4b5227a5bc0b5403b3054fc7fb81256ab0eef6" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10124", "title": "Training with Missed Targets in Generative Recommendation: Separating Supervision from Probability Competition", "published": "2026-10-07T14:07:06Z", "reviews": [ { "id": "2610.10124", "title": "Training with Missed Targets in Generative Recommendation: Separating Supervision from Probability Competition", "published": "2026-10-07T14:07:06Z", "decision": "reviewed_deferred", "reason": "独立作者/浙大对missed-target reranker loss做三组匹配分解，OneRec/Amazon公开离线结果及development-set选损失可作生成排序方法学参考。没有工业线上证据，不收为工业复现；跨领域loss研究留后续独立例外/通用训练审查，不在本批16项。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "bounded_methods_and_evaluation", "review_artifact": "research-audits/oct12-review-closure/recommendation-reviews.json", "evidence_locations": [ "§3 matched losses", "§4–§5 experiments", "§6 Discussion" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10124v1", "sha256": "2719045f0faecb2628363a3025498c0225d22b3637f96a0d43dfd21efc135523" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "recommendation" ], "status": "reviewed_deferred" }, { "id": "2610.10120", "title": "RewardWeaver: Long-Horizon Interactive Learning for Language Agents via Self-Evolving Reward Adaptation", "published": "2026-10-07T14:04:45Z", "reviews": [ { "id": "2610.10120", "title": "RewardWeaver: Long-Horizon Interactive Learning for Language Agents via Self-Evolving Reward Adaptation", "published": "2026-10-07T14:04:45Z", "decision": "reviewed_deferred", "reason": "已读RewardWeaver method/experiment：admitted rubric语义固定，阶段训练后lowoutcome backwardattribution动态reward allocation，独立controlled expansion；社会/议价/sales。下一批须实际policy训练、validatedrubric稳定，不能追分随意重定义reward或测试goal泄漏。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 RewardWeaver Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10120v1", "sha256": "aee6e3d6daa3eecf37e11853e79c80c3307d9312fbbac6db2113852d19aaa5d2", "local_html_sha256": "aee6e3d6daa3eecf37e11853e79c80c3307d9312fbbac6db2113852d19aaa5d2", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10120v1", "result": "retrieved", "sha256": "aee6e3d6daa3eecf37e11853e79c80c3307d9312fbbac6db2113852d19aaa5d2" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10118", "title": "YANchor-4B: Effective Long-Horizon Reasoning in O(N) Time with O(1) Memory", "published": "2026-10-07T14:03:50Z", "reviews": [ { "id": "2610.10118", "title": "YANchor-4B: Effective Long-Horizon Reasoning in O(N) Time with O(1) Memory", "published": "2026-10-07T14:03:50Z", "decision": "reviewed_deferred", "reason": "YANchor GDN+SWA1024/anchorstatebranch，27text6VL与H100throughput；需真memory训练，均值pass1重复抽样不混成bestofK。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3  Architecture", "location": ".cache/oct12-scan/fulltext/2610.10118.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5  Evaluation", "location": ".cache/oct12-scan/fulltext/2610.10118.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 191, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10118v1", "sha256": "fbc03ddb6be01e6f5a381ca1039bf033fa3df20a83a17a8fd34deb4cc9fb468e", "local_path": ".cache/oct12-scan/fulltext/2610.10118.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10118v1", "result": "retrieved", "sha256": "fbc03ddb6be01e6f5a381ca1039bf033fa3df20a83a17a8fd34deb4cc9fb468e" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10118", "sha256": "880a0e2d5bdb62740a72bf062d2df69f537a2d8fbaf6cb74bcac9c83a086fd9d", "local_path": ".cache/oct12-scan/fulltext/2610.10118.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 14:03:50 UTC (1,480 KB)", "published": "2026-10-07T14:03:50Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10114", "title": "Mechanics of Long-Context Hybrid Models Part 1.1: From Hybrid Attention to Hybrid Position", "published": "2026-10-07T14:02:35Z", "reviews": [ { "id": "2610.10114", "title": "Mechanics of Long-Context Hybrid Models Part 1.1: From Hybrid Attention to Hybrid Position", "published": "2026-10-07T14:02:35Z", "decision": "reviewed_deferred", "reason": "hybridposition：50Bshort/5Blong多模型，PG19/RULER/BABILong；SWA/LA seesaw与slidingLA，单NIAH100%不外推全部长上下文。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "additional targeted method/evaluation", "heading": "3 Observation", "section_index": 2, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "6 Discussion on LA Hybrid", "section_index": 5, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 192, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10114v1", "sha256": "bac5b490398d94ce25fdd2baabead88cddec4deb05ba2f9744b596aec3a6e932", "local_path": ".cache/oct12-scan/fulltext/2610.10114.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10114v1", "result": "retrieved", "sha256": "bac5b490398d94ce25fdd2baabead88cddec4deb05ba2f9744b596aec3a6e932" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10114", "sha256": "cd4bbc7f26ea3b23074bb4d259d9500ea7b1f47cda152f09f8aaaede792c7218", "local_path": ".cache/oct12-scan/fulltext/2610.10114.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 14:02:35 UTC (4,805 KB)", "published": "2026-10-07T14:02:35Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10092", "title": "I would rather quit NLP than read another paper like this: The rise of antithesis in NLP papers", "published": "2026-10-07T13:50:50Z", "reviews": [ { "id": "2610.10092", "title": "I would rather quit NLP than read another paper like this: The rise of antithesis in NLP papers", "published": "2026-10-07T13:50:50Z", "decision": "rejected_or_out_of_scope", "reason": "ACL/arXiv写作antithesis频率、双annotator和rewardmodel偏好研究，posttraining只作机制猜测；无具体训练算法可实施，当前batch outofscope，可作为写作偏差背景。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Data", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Results", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10092v1", "sha256": "16de4adb67313eced254c8e2f2047a9e46b1aecb275bc28bdff469a58efe9715", "local_html_sha256": "16de4adb67313eced254c8e2f2047a9e46b1aecb275bc28bdff469a58efe9715", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10092v1", "result": "retrieved", "sha256": "16de4adb67313eced254c8e2f2047a9e46b1aecb275bc28bdff469a58efe9715" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10091", "title": "ExperienceIndex: Artifact-Grounded Memory", "published": "2026-10-07T13:50:33Z", "reviews": [ { "id": "2610.10091", "title": "ExperienceIndex: Artifact-Grounded Memory", "published": "2026-10-07T13:50:33Z", "decision": "reviewed_deferred", "reason": "已读ExperienceIndex方法/实验：真实priortrace提artifact contribution及pairrelations、retrieve指导新任务artifactset，跨taskteacher/student。下一批需要可溯源memory形成、evaluation task答案和目标issue未来fix不能进入experience。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 ExperienceIndex", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10091v1", "sha256": "9d539dc4db6e33c4909a23291ec47e925f4aea23a2bff1dc7e5f2b9801988a35", "local_html_sha256": "9d539dc4db6e33c4909a23291ec47e925f4aea23a2bff1dc7e5f2b9801988a35", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10091v1", "result": "retrieved", "sha256": "9d539dc4db6e33c4909a23291ec47e925f4aea23a2bff1dc7e5f2b9801988a35" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10071", "title": "HGP:An on-device personalized agent memory via hybrid graph storage", "published": "2026-10-07T13:37:38Z", "reviews": [ { "id": "2610.10071", "title": "HGP:An on-device personalized agent memory via hybrid graph storage", "published": "2026-10-07T13:37:38Z", "decision": "reviewed_deferred", "reason": "已读HGP方法/实验：confidence-driven LLMannotation在线训lightweightrouter、episode/semantic/proceduralgraphs+workingstate；LoCoMo/PAL选择与QA。下一批须真实router更新与图检索、on-device latency/资源实际测，不能只标签分类fixture。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10071v1", "sha256": "5ffbdfc6a6f284cb5bb73466695f17425f0f84007ffd4ea82b97b6f28dad7b4b", "local_html_sha256": "5ffbdfc6a6f284cb5bb73466695f17425f0f84007ffd4ea82b97b6f28dad7b4b", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10071v1", "result": "retrieved", "sha256": "5ffbdfc6a6f284cb5bb73466695f17425f0f84007ffd4ea82b97b6f28dad7b4b" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10068", "title": "Efficient Provably Private Classification with a Tabular Foundation Model", "published": "2026-10-07T13:34:38Z", "reviews": [ { "id": "2610.10068", "title": "Efficient Provably Private Classification with a Tabular Foundation Model", "published": "2026-10-07T13:34:38Z", "decision": "reviewed_deferred", "reason": "PrivTab敏感context变DPsummary、simulatedpretrain，33TabArena10splits；formalGDP/sensitivity/预处理会计需实现，低MI不能替代保证。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "Non-private models reveal membership", "location": ".cache/oct12-scan/fulltext/2610.10068.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "Principled and efficient privacy-aware in-context learning", "section_index": 0, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "Private classification across datasets", "section_index": 1, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "Methods", "section_index": 5, "characters_read": [ 0, 1422 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 193, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10068v1", "sha256": "674323db338a9906811035168848ee25845d6eaf22b5a46c9d5f0b67d53e621a", "local_path": ".cache/oct12-scan/fulltext/2610.10068.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10068v1", "result": "retrieved", "sha256": "674323db338a9906811035168848ee25845d6eaf22b5a46c9d5f0b67d53e621a" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10068", "sha256": "6ec472e3d6e3352d620044ea9b4308a9b5235295ce506fe390cd421b6ae487d0", "local_path": ".cache/oct12-scan/fulltext/2610.10068.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 13:34:38 UTC (26,853 KB)", "published": "2026-10-07T13:34:38Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10066", "title": "From Pixel to Coding: Evaluating the Figure Reproduction Capabilities of MLLMs", "published": "2026-10-07T13:33:04Z", "reviews": [ { "id": "2610.10066", "title": "From Pixel to Coding: Evaluating the Figure Reproduction Capabilities of MLLMs", "published": "2026-10-07T13:33:04Z", "decision": "reviewed_deferred", "reason": "FigCodeBench6k/4programlanguage复杂度，24MLLM真实render视觉+syntaxisomorphism；需编译执行与humanalign评分，不用字符串替代。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "3 Benchmark Curation", "location": ".cache/oct12-scan/fulltext/2610.10066.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "4 Experiments", "section_index": 3, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 194, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10066v1", "sha256": "78f5a68fadc2373dff90253f37a64354894cd7365ea0d11ea0e4a143e9f727b8", "local_path": ".cache/oct12-scan/fulltext/2610.10066.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10066v1", "result": "retrieved", "sha256": "78f5a68fadc2373dff90253f37a64354894cd7365ea0d11ea0e4a143e9f727b8" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10066", "sha256": "6989ab5661bd89fc955393771058719ef742b19e85e914d087afe8439dd5e2e6", "local_path": ".cache/oct12-scan/fulltext/2610.10066.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 13:33:04 UTC (21,571 KB)", "published": "2026-10-07T13:33:04Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10062", "title": "Loud Failures, Quiet Failures: Fault Detection and Recovery in Tool-Using Language Model Agents", "published": "2026-10-07T13:29:59Z", "reviews": [ { "id": "2610.10062", "title": "Loud Failures, Quiet Failures: Fault Detection and Recovery in Tool-Using Language Model Agents", "published": "2026-10-07T13:29:59Z", "decision": "reviewed_deferred", "reason": "已读loud/quietfailure method/results：自有BFCL statefultool layer四typedfault控制注入，24task1920trial、faultfree nondeterminism基线与检测/改plan/recovery分开。下一批须真实tool状态execute，gold finalstate仅scorer，reasoning无recovery提升负结果保留。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "III Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "IV Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10062v1", "sha256": "3af91fd44af6986f7665aeef63d2b62da03596c2b6c2d38ef63aac99e40eb1ea", "local_html_sha256": "3af91fd44af6986f7665aeef63d2b62da03596c2b6c2d38ef63aac99e40eb1ea", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10062v1", "result": "retrieved", "sha256": "3af91fd44af6986f7665aeef63d2b62da03596c2b6c2d38ef63aac99e40eb1ea" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10061", "title": "TRACK: Telemetry-Based Racing Analysis and Coaching Kit in Sim Racing Games", "published": "2026-10-07T13:29:45Z", "reviews": [ { "id": "2610.10061", "title": "TRACK: Telemetry-Based Racing Analysis and Coaching Kit in Sim Racing Games", "published": "2026-10-07T13:29:45Z", "decision": "rejected_or_out_of_scope", "reason": "已读完整TRACK摘要，官方正文未取得：simracing telemetry四维fingerprint与无监督clustering、RL仅固定reference尺度，非LLMagent/新policy算法；小样本且crosscarrepeatability未证，当前范围外。", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "read_scope": "abstract_only_primary_fulltext_unavailable", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "url": "https://arxiv.org/abs/2610.10061", "sha256": null, "local_html_sha256": null, "local_html_sha256_matches": false, "attempts": [ { "url": "https://arxiv.org/html/2610.10061v1", "result": "failed", "error": "404 Client Error: Not Found for url: https://arxiv.org/html/2610.10061v1" }, { "url": "https://arxiv.org/html/2610.10061v1", "result": "failed", "error": "404 Client Error: Not Found for url: https://arxiv.org/html/2610.10061v1" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10058", "title": "Cache the Encoder Within:Compact, Reusable Memory across LLM Queries", "published": "2026-10-07T13:28:25Z", "reviews": [ { "id": "2610.10058", "title": "Cache the Encoder Within:Compact, Reusable Memory across LLM Queries", "published": "2026-10-07T13:28:25Z", "decision": "reviewed_deferred", "reason": "EncBank cachedlowerstates+sharedselfdistilledsuffix，三Qwen五suite；native/precision不同checkpoint不叠加收益，write/fetch/decode成本另算。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "2 Design Insights for Reusable Memory", "location": ".cache/oct12-scan/fulltext/2610.10058.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experimental Design", "location": ".cache/oct12-scan/fulltext/2610.10058.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 Methodology", "section_index": 2, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 195, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10058v1", "sha256": "8445591b6a362eeb6b6958333c2ba5f08f0cb6e7058e65a858d360959177a756", "local_path": ".cache/oct12-scan/fulltext/2610.10058.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10058v1", "result": "retrieved", "sha256": "8445591b6a362eeb6b6958333c2ba5f08f0cb6e7058e65a858d360959177a756" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10058", "sha256": "e539e69f9a7cb8997488a77c1f8ba946254aef939eb284ad868b97755334af3b", "local_path": ".cache/oct12-scan/fulltext/2610.10058.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 13:28:25 UTC (89 KB)", "published": "2026-10-07T13:28:25Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10633", "title": "Has LLM Screening Performance Stalled in Software Engineering Systematic Reviews?", "published": "2026-10-07T13:25:54Z", "reviews": [ { "id": "2610.10633", "title": "Has LLM Screening Performance Stalled in Software Engineering Systematic Reviews?", "published": "2026-10-07T13:25:54Z", "decision": "reviewed_deferred", "reason": "已读LLM筛paper method/results：SESR-Eval/Mini sampling、多criterion/MCC与lostevidence，新模型小幅变化、agreement不等正确。agent-based筛选仅futureproposal，保留评测参考不宣称有agent实现，也不能替代本次人工审查。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10633v1", "sha256": "701872741619249616cefcb0691d0beaa95d9f949b35aec56046f841fd0d0813", "local_html_sha256": "701872741619249616cefcb0691d0beaa95d9f949b35aec56046f841fd0d0813", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10633v1", "result": "retrieved", "sha256": "701872741619249616cefcb0691d0beaa95d9f949b35aec56046f841fd0d0813" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10042", "title": "Learning to Accumulate Knowledge with Mutual Information", "published": "2026-10-07T13:20:18Z", "reviews": [ { "id": "2610.10042", "title": "Learning to Accumulate Knowledge with Mutual Information", "published": "2026-10-07T13:20:18Z", "decision": "reviewed_deferred", "reason": "已读KnowledgeWeaver method/experiment：实际traincurator、frozenexecutor，tokenMI类feedback+standalone与marginalsuccess，bank增量减少redundancy；ALFWorld/WebShop。下一批需真GRPO与执行reward、bankheldout隔离，固定knowledge/goldplan非核心复现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10042v1", "sha256": "60166d360ef7c9e1f3125be192502b2cd7e80ce7746578a2e8bd1474b3adcc64", "local_html_sha256": "60166d360ef7c9e1f3125be192502b2cd7e80ce7746578a2e8bd1474b3adcc64", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10042v1", "result": "retrieved", "sha256": "60166d360ef7c9e1f3125be192502b2cd7e80ce7746578a2e8bd1474b3adcc64" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09995", "title": "Designing Collaborative AI-Driven Workflows for Scientific Software Engineering", "published": "2026-10-07T12:55:07Z", "reviews": [ { "id": "2610.09995", "title": "Designing Collaborative AI-Driven Workflows for Scientific Software Engineering", "published": "2026-10-07T12:55:07Z", "decision": "reviewed_deferred", "reason": "已读scientificworkflow架构/MCFMcase：domainexpert定义spec/阶段numericreference与humanapproval，Fortran->C++14realbuild比较author-reviewloop cost。下一批真实编译/numericalvalidation，不能论文human-inloop结果宣传全自治能力。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Case Study: MCFM", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 Collaborative Workflow Architecture", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09995v1", "sha256": "33266ee917654c3834d00b0b8a5ee3e7708128c4394446a3cf9ccad2378e53ff", "local_html_sha256": "33266ee917654c3834d00b0b8a5ee3e7708128c4394446a3cf9ccad2378e53ff", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09995v1", "result": "retrieved", "sha256": "33266ee917654c3834d00b0b8a5ee3e7708128c4394446a3cf9ccad2378e53ff" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09985", "title": "Marrying Pricing and Advertising with LLMs", "published": "2026-10-07T12:50:08Z", "reviews": [ { "id": "2610.09985", "title": "Marrying Pricing and Advertising with LLMs", "published": "2026-10-07T12:50:08Z", "decision": "reviewed_deferred", "reason": "LLMLoRA广告actor/需求critic与REINFORCEbaseline联合pricing，synthetic和marketplacefittedsim；expectedrevenue非生产AB。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "4 Experimental Results", "location": ".cache/oct12-scan/fulltext/2610.09985.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 Online actor-critic algorithm", "section_index": 2, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 196, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09985v1", "sha256": "d1662516a5682f6f87970d3aa197256286dec5f75188abccc2961177e68440dc", "local_path": ".cache/oct12-scan/fulltext/2610.09985.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09985v1", "result": "retrieved", "sha256": "d1662516a5682f6f87970d3aa197256286dec5f75188abccc2961177e68440dc" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09985", "sha256": "a895ecec735eea2011a1c2496c2b999c40a6487e7ad12895118d38639e7219c1", "local_path": ".cache/oct12-scan/fulltext/2610.09985.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 12:50:08 UTC (1,372 KB)", "published": "2026-10-07T12:50:08Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09973", "title": "From Expected Harmfulness to Likelihood: A Probabilistic Reformulation of Jailbreaking LLM Agents", "published": "2026-10-07T12:40:20Z", "reviews": [ { "id": "2610.09973", "title": "From Expected Harmfulness to Likelihood: A Probabilistic Reformulation of Jailbreaking LLM Agents", "published": "2026-10-07T12:40:20Z", "decision": "reviewed_deferred", "reason": "已读OPUR expectedharm理论/实验高层：harm reweighted outputsampling与likelihood objective联系，AgentHarm等安全benchmark。保留安全研究候选，不执行未授权服务jailbreak/ harmfultool工作流，后续须严格sandbox threatmodel。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "5 EXPERIMENT", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 PRELIMINAIRES", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09973v1", "sha256": "0bc3681a9849139c9cbe20f9c04c365f384dfdb61f8b9083dc8650f13ddc5d9f", "local_html_sha256": "0bc3681a9849139c9cbe20f9c04c365f384dfdb61f8b9083dc8650f13ddc5d9f", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09973v1", "result": "retrieved", "sha256": "0bc3681a9849139c9cbe20f9c04c365f384dfdb61f8b9083dc8650f13ddc5d9f" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09969", "title": "TR-PTQ: High-Accuracy Integer-Only Transformer Post Training Quantization via Taylor Region Reformulation", "published": "2026-10-07T12:39:10Z", "reviews": [ { "id": "2610.09969", "title": "TR-PTQ: High-Accuracy Integer-Only Transformer Post Training Quantization via Taylor Region Reformulation", "published": "2026-10-07T12:39:10Z", "decision": "reviewed_deferred", "reason": "TaylorRegion整数exp/log与LayerNorm/GELU/Softmax，vision/languagePTQ；需要真实integerkernel及同FPbaseline，不能fakequant等同硬件部署。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "5 Experiments and Evaluation", "location": ".cache/oct12-scan/fulltext/2610.09969.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 Integer-only Taylor Region Approximation for Nonlinear Functions", "section_index": 2, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 197, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09969v1", "sha256": "a0546e522dda7dfed857b83d4feb4b883d64ce499ae1af95501dbc04fe9e50c5", "local_path": ".cache/oct12-scan/fulltext/2610.09969.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09969v1", "result": "retrieved", "sha256": "a0546e522dda7dfed857b83d4feb4b883d64ce499ae1af95501dbc04fe9e50c5" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09969", "sha256": "21c1acd54d8b303ab98988512119bf8b8fc3ecbcb336a3915c27838fd7b88d12", "local_path": ".cache/oct12-scan/fulltext/2610.09969.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 12:39:10 UTC (1,111 KB)", "published": "2026-10-07T12:39:10Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10629", "title": "The Harness as the Only Mutable Surface: Compliance-Bounded Self-Evolution of LLM Agents in Credit Pipelines, with a Measured Admission Gate", "published": "2026-10-07T12:33:48Z", "reviews": [ { "id": "2610.10629", "title": "The Harness as the Only Mutable Surface: Compliance-Bounded Self-Evolution of LLM Agents in Credit Pipelines, with a Measured Admission Gate", "published": "2026-10-07T12:33:48Z", "decision": "reviewed_deferred", "reason": "已读credit gate threatmodel/results：simulatedagent和seeded-search非LM，harness唯一mutable、hashchainadmission、防仅recenterror促成放松screen；高severitygate误拒正确替换。仅机制simulationdiagnostic，无真实LLM训练或生产validation，本批不作能力候选。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Problem, threat model and requirements", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "7 Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10629v1", "sha256": "aa77c2b5180a54a5ba883158f2f493e3343c76b156087ba44175c9f54233f176", "local_html_sha256": "aa77c2b5180a54a5ba883158f2f493e3343c76b156087ba44175c9f54233f176", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10629v1", "result": "retrieved", "sha256": "aa77c2b5180a54a5ba883158f2f493e3343c76b156087ba44175c9f54233f176" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09952", "title": "MSU Team at the Explainable Deepfake Detection Challenge 2026: Grounded Artifact Evidence for Deepfake Detection", "published": "2026-10-07T12:30:04Z", "reviews": [ { "id": "2610.09952", "title": "MSU Team at the Explainable Deepfake Detection Challenge 2026: Grounded Artifact Evidence for Deepfake Detection", "published": "2026-10-07T12:30:04Z", "decision": "reviewed_deferred", "reason": "XPlainVerse DINO/Mesorchdetector、trainingartifactmask和patchcontrastive+QwenVL解释/GRPO简化；publictrainval/test可验，真多stage待实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "4. Methodology", "location": ".cache/oct12-scan/fulltext/2610.09952.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5. Experiments", "location": ".cache/oct12-scan/fulltext/2610.09952.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 198, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09952v1", "sha256": "c22d1973f3cf36a9eccc6d720caf92253acdce869d9aaa9feb450c9dab41818b", "local_path": ".cache/oct12-scan/fulltext/2610.09952.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09952v1", "result": "retrieved", "sha256": "c22d1973f3cf36a9eccc6d720caf92253acdce869d9aaa9feb450c9dab41818b" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09952", "sha256": "7ec291fe511ba0d0bc51cfee8f2f44496bd709b069754c0cddf52f127d133628", "local_path": ".cache/oct12-scan/fulltext/2610.09952.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 12:30:04 UTC (2,381 KB)", "published": "2026-10-07T12:30:04Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09941", "title": "Purifying Backdoored Large Vision-Language Models by Removing Hijacked Directions", "published": "2026-10-07T12:25:47Z", "reviews": [ { "id": "2610.09941", "title": "Purifying Backdoored Large Vision-Language Models by Removing Hijacked Directions", "published": "2026-10-07T12:25:47Z", "decision": "reviewed_deferred", "reason": "OrthoPurify smallclean伪benignweight方向识别并单次orthogonalprojection，LLaVA/Qwen六backdoor/COCO/VQAv2；真实cleantrain与adaptive威胁待核。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Our Method: OrthoPurify", "location": ".cache/oct12-scan/fulltext/2610.09941.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09941.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 199, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09941v1", "sha256": "89bfd0d4fbdfd4cef3bf922715e75a50baf20b52cc32fa6e67476ad50cb33e05", "local_path": ".cache/oct12-scan/fulltext/2610.09941.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09941v1", "result": "retrieved", "sha256": "89bfd0d4fbdfd4cef3bf922715e75a50baf20b52cc32fa6e67476ad50cb33e05" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09941", "sha256": "2dcea3348955647aaccdeea7c54464c538b88f8b32bbeb3c3d422d95bdaa9366", "local_path": ".cache/oct12-scan/fulltext/2610.09941.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 12:25:47 UTC (3,011 KB)", "published": "2026-10-07T12:25:47Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09940", "title": "Juno: Taming Predictive Latents for Vision-Language-Action Models", "published": "2026-10-07T12:24:17Z", "reviews": [ { "id": "2610.09940", "title": "Juno: Taming Predictive Latents for Vision-Language-Action Models", "published": "2026-10-07T12:24:17Z", "decision": "reviewed_deferred", "reason": "Juno actionJEPA动态CLS、decoupledfuturelatentdistill，部署observedtransitions适配后verifiedexecution对齐；真实robot/shift交互与testtime训练必须分开。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Method", "location": ".cache/oct12-scan/fulltext/2610.09940.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09940.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 200, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09940v1", "sha256": "ea8aff08fd41f7bb7f65129078a98c3ecdbd6acaecfb58d2aa0f6b344560349b", "local_path": ".cache/oct12-scan/fulltext/2610.09940.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09940v1", "result": "retrieved", "sha256": "ea8aff08fd41f7bb7f65129078a98c3ecdbd6acaecfb58d2aa0f6b344560349b" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09940", "sha256": "9347ef89f73890186114967af8912731b8cffab929cd7f6bdf77e0a5567a6ef9", "local_path": ".cache/oct12-scan/fulltext/2610.09940.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 12:24:17 UTC (12,963 KB)", "published": "2026-10-07T12:24:17Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09935", "title": "AgentTracer: Tracing Indirect Prompt Injection Attack through Fine-Grained Intention-Execution Alignment", "published": "2026-10-07T12:19:56Z", "reviews": [ { "id": "2610.09935", "title": "AgentTracer: Tracing Indirect Prompt Injection Attack through Fine-Grained Intention-Execution Alignment", "published": "2026-10-07T12:19:56Z", "decision": "reviewed_deferred", "reason": "已读AgentTracer graphdesign/evaluation：userintent authorization与operationKB、taskintent implicitdependency、从anomalouscall backwardprune追injectionsource；successattacklogs加背景noise。下一批须完整trace与无attackfalsepositive控制，不能已知injection位置当attributor输入。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4. Design Details of AgentTracer", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5. Evaluation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09935v1", "sha256": "1ea45b7c4c3ee80cc4811ae8e4ab2788cc2de10db020e303ff1d8b72ecb70da7", "local_html_sha256": "1ea45b7c4c3ee80cc4811ae8e4ab2788cc2de10db020e303ff1d8b72ecb70da7", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09935v1", "result": "retrieved", "sha256": "1ea45b7c4c3ee80cc4811ae8e4ab2788cc2de10db020e303ff1d8b72ecb70da7" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09934", "title": "Itgan at NADI 2026 shared task: Parameter-Efficient Whisper Adaptation for Robust, Mixed-Dialect and Code-Switched Arabic ASR", "published": "2026-10-07T12:18:40Z", "reviews": [ { "id": "2610.09934", "title": "Itgan at NADI 2026 shared task: Parameter-Efficient Whisper Adaptation for Robust, Mixed-Dialect and Code-Switched Arabic ASR", "published": "2026-10-07T12:18:40Z", "decision": "reviewed_deferred", "reason": "NADI WhisperwideLoRA Arabicdialect/ code-switch ASR、specialistrouting/weightspaceaverage/ROVER有bootstrap与negative results；foundation speechadaptation下一批，oracle dialect routing需标注不可用label的postevalprobe。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Common Setup", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Subtask 1.1: Robust Country-Level ASR", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09934v1", "sha256": "f2ceccf28e1f629e8596b3a89ad83598551bfcda3fe2a246267f3f2d4d0f95ec", "local_html_sha256": "f2ceccf28e1f629e8596b3a89ad83598551bfcda3fe2a246267f3f2d4d0f95ec", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09934v1", "result": "retrieved", "sha256": "f2ceccf28e1f629e8596b3a89ad83598551bfcda3fe2a246267f3f2d4d0f95ec" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09920", "title": "Inverting Multi-Vector Visual Document Indices", "published": "2026-10-07T12:10:11Z", "reviews": [ { "id": "2610.09920", "title": "Inverting Multi-Vector Visual Document Indices", "published": "2026-10-07T12:10:11Z", "decision": "reviewed_deferred", "reason": "visualdocument index条件image inversion与orderrestore，ViDoRev3/跨retriever；OCR敏感token恢复与sourcepagerank非同metric，真生成模型待核。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Threat Model and Problem", "location": ".cache/oct12-scan/fulltext/2610.09920.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5 Experimental Setup", "location": ".cache/oct12-scan/fulltext/2610.09920.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 201, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09920v1", "sha256": "c985a8740a60e0149ffe93fb0bdad1035cb14d78e419711a7fd494246decb12a", "local_path": ".cache/oct12-scan/fulltext/2610.09920.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09920v1", "result": "retrieved", "sha256": "c985a8740a60e0149ffe93fb0bdad1035cb14d78e419711a7fd494246decb12a" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09920", "sha256": "94299cad455086a408552f1f4e2985be8167db0c00f550c2f43d4e5085aa6afd", "local_path": ".cache/oct12-scan/fulltext/2610.09920.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 12:10:11 UTC (16,687 KB)", "published": "2026-10-07T12:10:11Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09919", "title": "Eigenvalues of the Hessian in Deep Learning: The Origin of Symmetry and Its Breaking", "published": "2026-10-07T12:08:48Z", "reviews": [ { "id": "2610.09919", "title": "Eigenvalues of the Hessian in Deep Learning: The Origin of Symmetry and Its Breaking", "published": "2026-10-07T12:08:48Z", "decision": "reviewed_deferred", "reason": "Hessian symmetrybreaking理论解释kernels/clusters，对多架构spectralanalysis；无可验证foundation任务收益/部署算例，本批defer理论而非能力实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Concepts and Methods", "location": ".cache/oct12-scan/fulltext/2610.09919.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "4 Spectral Analysis", "section_index": 3, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 202, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09919v1", "sha256": "50298c0c89408d1fb7f4d83a94153d5aea20d38375830bf51693ef757bc43f7a", "local_path": ".cache/oct12-scan/fulltext/2610.09919.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09919v1", "result": "retrieved", "sha256": "50298c0c89408d1fb7f4d83a94153d5aea20d38375830bf51693ef757bc43f7a" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09919", "sha256": "2d2eee14a4403b1f1a60c7f594cb7f15a899fc1f95ff25b494946d5c17d07157", "local_path": ".cache/oct12-scan/fulltext/2610.09919.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 12:08:48 UTC (1,857 KB)", "published": "2026-10-07T12:08:48Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09914", "title": "RollVerify: Bridging Efficiency and Accuracy in Long-Tail Rollout Reinforcement Learning", "published": "2026-10-07T12:05:33Z", "reviews": [ { "id": "2610.09914", "title": "RollVerify: Bridging Efficiency and Accuracy in Long-Tail Rollout Reinforcement Learning", "published": "2026-10-07T12:05:33Z", "decision": "reviewed_deferred", "reason": "RollVerify以OPS衡量partial rollout off-policy偏移，并在训练前sequence/token验证截断repair suffix；VeRL Qwen3-8B math/tool-math及初步code对照，下一批需真实stale sampling/repair和throughput，非仅importance loss。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "4 RollVerify", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09914v1", "sha256": "4c34b39e848c01e4f41107803340bcbca7458df59abed5858d423ef8e62c7035", "local_html_sha256": "4c34b39e848c01e4f41107803340bcbca7458df59abed5858d423ef8e62c7035", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09914v1", "result": "retrieved", "sha256": "4c34b39e848c01e4f41107803340bcbca7458df59abed5858d423ef8e62c7035" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09907", "title": "FedSSMCoOp: SSM Encoders for light-weight Federated Prompt Learning for Few-shot Classification", "published": "2026-10-07T12:00:59Z", "reviews": [ { "id": "2610.09907", "title": "FedSSMCoOp: SSM Encoders for light-weight Federated Prompt Learning for Few-shot Classification", "published": "2026-10-07T12:00:59Z", "decision": "reviewed_deferred", "reason": "FedSSMCoOp Vision/CrossMamba与federatedsoft/communicationprompt，11biomedicalbench；federated并非自动privacy保证，code未release/GPU待核。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "4 Proposed Approach", "location": ".cache/oct12-scan/fulltext/2610.09907.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09907.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 203, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09907v1", "sha256": "c0416de8075086ba032c34c669bac83ffac2115236f8ffdedf692a2edb5b9802", "local_path": ".cache/oct12-scan/fulltext/2610.09907.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09907v1", "result": "retrieved", "sha256": "c0416de8075086ba032c34c669bac83ffac2115236f8ffdedf692a2edb5b9802" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09907", "sha256": "65dc1326d4c1c7e4fb3187f0378c8953881a23fcf61a95dd77a69b5c278de8ed", "local_path": ".cache/oct12-scan/fulltext/2610.09907.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 12:00:59 UTC (1,386 KB)", "published": "2026-10-07T12:00:59Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09906", "title": "Constrained-Action AI Remediation for SIEM/XDR via a NeMo-Guardrails Proxy", "published": "2026-10-07T12:00:51Z", "reviews": [ { "id": "2610.09906", "title": "Constrained-Action AI Remediation for SIEM/XDR via a NeMo-Guardrails Proxy", "published": "2026-10-07T12:00:51Z", "decision": "reviewed_deferred", "reason": "SIEM/XDR闭集intent与参数验证+现成NeMo proxy的安全评测；§VI 200攻击/1000良性及真实红队，§VII称OT仅架构适配非实测部署。与Agent安全相关，需公开corpus/完整代理边界复核和隔离沙箱，转下一批；不授权真实端点remediation。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "bounded_methods_and_evaluation", "review_artifact": "research-audits/oct12-review-closure/recommendation-reviews.json", "evidence_locations": [ "§III–§V", "§VI Preliminary Evaluation", "§VII Limitations" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09906v1", "sha256": "4c8de66b15a586c77fe9595cd945eff9aedea5c673cf21ae6e370322a21baba2" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model", "recommendation" ], "status": "reviewed_deferred" }, { "id": "2610.09872", "title": "LiveMACE: Process-Aware Evaluation of LLM Agent Capabilities in Evolving Markets", "published": "2026-10-07T11:33:21Z", "reviews": [ { "id": "2610.09872", "title": "LiveMACE: Process-Aware Evaluation of LLM Agent Capabilities in Evolving Markets", "published": "2026-10-07T11:33:21Z", "decision": "reviewed_deferred", "reason": "已读LiveMACE设计/评测：30日live market paper-trading，matchedtool/memory/rule/collaboration与mechanismtrace诊断，结果收益不等capability。下一批可只安全历史/纸交易，不执行资金操作；live quote流非真实生产投资A/B。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 LiveMACEBENCH : Benchmark Design", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Evaluation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09872v1", "sha256": "4a65b9940a70d76fc0eb7efa2079d03441344dc01599d9526d7b1efab42785b9", "local_html_sha256": "4a65b9940a70d76fc0eb7efa2079d03441344dc01599d9526d7b1efab42785b9", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09872v1", "result": "retrieved", "sha256": "4a65b9940a70d76fc0eb7efa2079d03441344dc01599d9526d7b1efab42785b9" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09870", "title": "Deadline-Aware Multi-Agent Reinforcement Learning for TSN-Based Vehicular Edge Networks", "published": "2026-10-07T11:30:19Z", "reviews": [ { "id": "2610.09870", "title": "Deadline-Aware Multi-Agent Reinforcement Learning for TSN-Based Vehicular Edge Networks", "published": "2026-10-07T11:30:19Z", "decision": "rejected_or_out_of_scope", "reason": "已读TSN VEC system/实验：每queue MAPPO动态order/timeslot、车辆速度deadline，传统MARL网络调度非LLM agent，本批范围外；真实mobilitydata不等网络真实部署AB。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "III System Model and Problem Formulation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "V Performance Evaluation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09870v1", "sha256": "0ce32444f1ad31f4b09d5304245ba0daeab349b57d8f15b1b336660a0e7c6bd6", "local_html_sha256": "0ce32444f1ad31f4b09d5304245ba0daeab349b57d8f15b1b336660a0e7c6bd6", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09870v1", "result": "retrieved", "sha256": "0ce32444f1ad31f4b09d5304245ba0daeab349b57d8f15b1b336660a0e7c6bd6" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.10628", "title": "Agent4RE: A Self-Refining Multi-agent Framework for End-to-End Software Requirements Engineering and Benchmarking", "published": "2026-10-07T11:23:11Z", "reviews": [ { "id": "2610.10628", "title": "Agent4RE: A Self-Refining Multi-agent Framework for End-to-End Software Requirements Engineering and Benchmarking", "published": "2026-10-07T11:23:11Z", "decision": "reviewed_deferred", "reason": "已读Agent4RE方法/RE-E2E实验：sequential、自修与humanfeedback三workflow，多agentrequirements生成，50humanwritten文件text/judge/manualmetrics。下一批需judge校准与requirementgroundtruth隔离，文本四分量表不当可执行software交付能力。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4. Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3. Agent4RE", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10628v1", "sha256": "8bb6d57ee5c9e409c2c40f07d52a1dc5a16924ab931fc1d988cdd16ef85dda48", "local_html_sha256": "8bb6d57ee5c9e409c2c40f07d52a1dc5a16924ab931fc1d988cdd16ef85dda48", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10628v1", "result": "retrieved", "sha256": "8bb6d57ee5c9e409c2c40f07d52a1dc5a16924ab931fc1d988cdd16ef85dda48" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09858", "title": "Training Advisors for LLM Agents from Task Outcomes", "published": "2026-10-07T11:14:26Z", "reviews": [ { "id": "2610.09858", "title": "Training Advisors for LLM Agents from Task Outcomes", "published": "2026-10-07T11:14:26Z", "decision": "reviewed_deferred", "reason": "Caddie冻结base agent、用最终outcome RL训练自然语言critic，MuSiQue/ALFWorld训练并跨base和DeepDive/tau3迁移；下一批需实际critic intervention和continuation outcomes，不能reference critique fixture。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "2 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09858v1", "sha256": "e2c933d0175bd4636ad8e2512000b7f0a217990953aee9a125b2523471d7ec42", "local_html_sha256": "e2c933d0175bd4636ad8e2512000b7f0a217990953aee9a125b2523471d7ec42", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09858v1", "result": "retrieved", "sha256": "e2c933d0175bd4636ad8e2512000b7f0a217990953aee9a125b2523471d7ec42" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09856", "title": "Self-Evolve With a Reference:Anchored Training of Tool-Integrated Agents", "published": "2026-10-07T11:13:35Z", "reviews": [ { "id": "2610.09856", "title": "Self-Evolve With a Reference:Anchored Training of Tool-Integrated Agents", "published": "2026-10-07T11:13:35Z", "decision": "reviewed_deferred", "reason": "已读AnchorLoop方法/实验：历史Executor frozenanchor同时给cross-referenceadvantage和Curriculum一致性proxy，VeRL sandbox+交替真实训练；curriculum时anchor与executor同参数，不能冒称版本改善或correctness。下一批须保留dualagent训练和reference snapshot，不以selfconsistency当gold。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments and Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09856v1", "sha256": "094f0b05b57af652513da2fecb3855fca29d4ef17d5c31f4b3b39ac3674316e5", "local_html_sha256": "094f0b05b57af652513da2fecb3855fca29d4ef17d5c31f4b3b39ac3674316e5", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09856v1", "result": "retrieved", "sha256": "094f0b05b57af652513da2fecb3855fca29d4ef17d5c31f4b3b39ac3674316e5" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09841", "title": "ORCA: Hunting Compositional Failures in Text-to-Image Diffusion", "published": "2026-10-07T11:03:58Z", "reviews": [ { "id": "2610.09841", "title": "ORCA: Hunting Compositional Failures in Text-to-Image Diffusion", "published": "2026-10-07T11:03:58Z", "decision": "reviewed_deferred", "reason": "ORCA将T5-CLIP residual条件化orthogonal low-rank visual readout作为diffusion latent辅助loss；COCO DiT/U-ViT FID/GenEval训练对照，转foundation T2I下一批，必须真实diffusion/visual encoder与QR投影。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09841v1", "sha256": "cc643f18eee8de1d5c81929ff89816f48b202b92f66d68a7b76b68f9dcf7e57c", "local_html_sha256": "cc643f18eee8de1d5c81929ff89816f48b202b92f66d68a7b76b68f9dcf7e57c", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09841v1", "result": "retrieved", "sha256": "cc643f18eee8de1d5c81929ff89816f48b202b92f66d68a7b76b68f9dcf7e57c" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09835", "title": "A Deafening Silence: Catastrophic Forgetting Lives in the Output Embeddings of Tokens the Data Never Speaks", "published": "2026-10-07T11:01:01Z", "reviews": [ { "id": "2610.09835", "title": "A Deafening Silence: Catastrophic Forgetting Lives in the Output Embeddings of Tokens the Data Never Speaks", "published": "2026-10-07T11:01:01Z", "decision": "reviewed_deferred", "reason": "Absenttoken outputhead one-sidedsoftmaxgrad受Adamvhat放大，head-onlyepsilon提高防forgetting并可加replay；多scale/familycontrolledfreezing与posthoc失败显示需training-time，下一批真实outputembedding optimizergroup和riskcounts。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "5 The Intervention and Its Boundaries", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "5 The Intervention and Its Boundaries", "scope": "additional eight-setting intervention results excerpt", "characters_read_at_least": [ 1300, 3800 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09835v1", "sha256": "428548747a95624eaa45472690071a538d6d77eabe174d62f6b7348030990a3c", "local_html_sha256": "428548747a95624eaa45472690071a538d6d77eabe174d62f6b7348030990a3c", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09835v1", "result": "retrieved", "sha256": "428548747a95624eaa45472690071a538d6d77eabe174d62f6b7348030990a3c" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09832", "title": "SkillForge: Co-Evolving Skills and Agents via Dynamic Skill Lifecycles", "published": "2026-10-07T10:52:15Z", "reviews": [ { "id": "2610.09832", "title": "SkillForge: Co-Evolving Skills and Agents via Dynamic Skill Lifecycles", "published": "2026-10-07T10:52:15Z", "decision": "reviewed_deferred", "reason": "已读SkillForge摘要/GRPOprelim与experiment：pre-RL skillfitness预退休后SFT、真实RL同时trial/active/stable/retired及mutation，SkillFurnace5krecords。下一批需完整lifecycle policy训练、验证选择技能，staticlibrary/goldtaskskillfixture非机制。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "5 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Preliminaries", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09832v1", "sha256": "457f19bbd1b919e7f6b6713474198f903cf6dd8b29d28e757421a6e28f26d0a6", "local_html_sha256": "457f19bbd1b919e7f6b6713474198f903cf6dd8b29d28e757421a6e28f26d0a6", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09832v1", "result": "retrieved", "sha256": "457f19bbd1b919e7f6b6713474198f903cf6dd8b29d28e757421a6e28f26d0a6" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09824", "title": "Homogenization in Multi-Agent Systems", "published": "2026-10-07T10:49:29Z", "reviews": [ { "id": "2610.09824", "title": "Homogenization in Multi-Agent Systems", "published": "2026-10-07T10:49:29Z", "decision": "reviewed_deferred", "reason": "已读homogenization设定/应用评测：conformity/polarization/inertia在coding/hiring/peerreview多agent，stochasticity/mixedmodel不充分的负结果。下一批行为infra应测interaction机制与relatederrors，aggregateaccuracy或modelheterogeneity标签不证明diversity。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Homogenization in Multi-Agent Systems", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Evaluating Homogenization Across Applications", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09824v1", "sha256": "3d487f23eb73ce7cf6c2349941ea6839e2ac72c15884b11ec8a7314afe384ca8", "local_html_sha256": "3d487f23eb73ce7cf6c2349941ea6839e2ac72c15884b11ec8a7314afe384ca8", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09824v1", "result": "retrieved", "sha256": "3d487f23eb73ce7cf6c2349941ea6839e2ac72c15884b11ec8a7314afe384ca8" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09823", "title": "UltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image Generation", "published": "2026-10-07T10:47:52Z", "reviews": [ { "id": "2610.09823", "title": "UltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image Generation", "published": "2026-10-07T10:47:52Z", "decision": "reviewed_deferred", "reason": "UltraText432双语denseprompt/24configs，VLMjudge四维并coverage；valid-only得分有缺失选择、配置分辨率不同，真实render/judge/human验证待核。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 UltraText Bench: Benchmark Design", "location": ".cache/oct12-scan/fulltext/2610.09823.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "3 UltraText Bench: Benchmark Design", "location": ".cache/oct12-scan/fulltext/2610.09823.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "5 Experiments", "section_index": 4, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 204, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09823v1", "sha256": "8878c34edefaa7b92d85e3876c357f10a7c09edba956a65d270b873743838286", "local_path": ".cache/oct12-scan/fulltext/2610.09823.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09823v1", "result": "retrieved", "sha256": "8878c34edefaa7b92d85e3876c357f10a7c09edba956a65d270b873743838286" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09823", "sha256": "76091bee9d0cbdb491a927081b7316e9450a0835f1d83b57191ad8df54677c0a", "local_path": ".cache/oct12-scan/fulltext/2610.09823.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 10:47:52 UTC (43,061 KB)", "published": "2026-10-07T10:47:52Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09804", "title": "BoT-GRPO: Efficient Process-Reward RL for Reasoning via Bag-of-Token Aggregation", "published": "2026-10-07T10:23:17Z", "reviews": [ { "id": "2610.09804", "title": "BoT-GRPO: Efficient Process-Reward RL for Reasoning via Bag-of-Token Aggregation", "published": "2026-10-07T10:23:17Z", "decision": "reviewed_deferred", "reason": "BoT-GRPO按source-sequencelength逆权聚合tokenrewards/weightedgroupstats、criticfree；Reactcompile/VLMjudge和AIME对照，下一批需真实稳定tokenRM及compileexecutions，防cosmeticrewardhacking。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09804v1", "sha256": "7dfe9b575388eac9aab6d07d0f136e7952342d0c13f85bd9c20265d27806981a", "local_html_sha256": "7dfe9b575388eac9aab6d07d0f136e7952342d0c13f85bd9c20265d27806981a", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09804v1", "result": "retrieved", "sha256": "7dfe9b575388eac9aab6d07d0f136e7952342d0c13f85bd9c20265d27806981a" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09802", "title": "DisParQ: Self-Supervised Part Concepts for Interpretable Vision Foundation Models", "published": "2026-10-07T10:19:04Z", "reviews": [ { "id": "2610.09802", "title": "DisParQ: Self-Supervised Part Concepts for Interpretable Vision Foundation Models", "published": "2026-10-07T10:19:04Z", "decision": "reviewed_deferred", "reason": "DisParQ frozenDINO空间exclusiveprototype与quantizedattribute重建，七dataset/partretrieval；实际concept解码训练和标注隔离待下一批。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "4 Experimental evaluation", "location": ".cache/oct12-scan/fulltext/2610.09802.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 DisParQ", "section_index": 2, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 205, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09802v1", "sha256": "d37e63a0b00de96eaf3a1c979d4d9dd48c35bafb1c81ac680f644a74a41f1745", "local_path": ".cache/oct12-scan/fulltext/2610.09802.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09802v1", "result": "retrieved", "sha256": "d37e63a0b00de96eaf3a1c979d4d9dd48c35bafb1c81ac680f644a74a41f1745" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09802", "sha256": "36562faf5d8e33f1b7d960f69f559280b8bf9ef28a776f261e3f0256b86466a5", "local_path": ".cache/oct12-scan/fulltext/2610.09802.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 10:19:04 UTC (16,701 KB)", "published": "2026-10-07T10:19:04Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09795", "title": "MIRROR: From Imitation to Internalization in LLM Personalization", "published": "2026-10-07T10:10:01Z", "reviews": [ { "id": "2610.09795", "title": "MIRROR: From Imitation to Internalization in LLM Personalization", "published": "2026-10-07T10:10:01Z", "decision": "reviewed_deferred", "reason": "MIRROR reference-conditioned frozen selfteacher对studentrollouts分布蒸馏，MIRRORF informative-referencetoken focalanchor；LaMP/LongLaMP+OOD个性化，训练privilege合法但test不得读referencewording，下一批。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09795v1", "sha256": "dc63b1f4486fbe4e59a155691c3cacb6fe7ea43faaa96effa7af7fdbc3850c28", "local_html_sha256": "dc63b1f4486fbe4e59a155691c3cacb6fe7ea43faaa96effa7af7fdbc3850c28", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09795v1", "result": "retrieved", "sha256": "dc63b1f4486fbe4e59a155691c3cacb6fe7ea43faaa96effa7af7fdbc3850c28" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09788", "title": "Judging in Latent Space: Efficient Generative Reward Modeling via Semantics-Preserving Compression", "published": "2026-10-07T10:05:06Z", "reviews": [ { "id": "2610.09788", "title": "Judging in Latent Space: Efficient Generative Reward Modeling via Semantics-Preserving Compression", "published": "2026-10-07T10:05:06Z", "decision": "reviewed_deferred", "reason": "LatentGRM semantic-chunk compression/reconstruction训练连续latent judge，另interpreter重构criterion评测，matched4B/8B preference和latency控制；下一批需真实continuous steps不能短文本proxy。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09788v1", "sha256": "3def777973d069f686e65d1c5752fdc2519351c8ad456ed15bded5581905b51d", "local_html_sha256": "3def777973d069f686e65d1c5752fdc2519351c8ad456ed15bded5581905b51d", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09788v1", "result": "retrieved", "sha256": "3def777973d069f686e65d1c5752fdc2519351c8ad456ed15bded5581905b51d" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09772", "title": "Decoupling Logic from Persona: Structural Immunity of Edge LLM Agents to Context Pollution", "published": "2026-10-07T09:53:28Z", "reviews": [ { "id": "2610.09772", "title": "Decoupling Logic from Persona: Structural Immunity of Edge LLM Agents to Context Pollution", "published": "2026-10-07T09:53:28Z", "decision": "reviewed_deferred", "reason": "同INT4 base的What logic LoRA只读core/vitals，生成Micro-State；How persona adapter读history与structuredstate，deterministic router/cache状态变化失效；两次decode不是简单promptwordfilter。; M2、Llama8B/Gemma4B、两task两persona五seed pollution ablations；低pollution无逻辑优势，4B polluted-control崩溃。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.09772v1#S3", "https://arxiv.org/html/2610.09772v1#S4", "https://arxiv.org/html/2610.09772v1#S5", "https://arxiv.org/html/2610.09772v1#S8" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09772v1", "sha256": "2664e6eff1883971bd197cf01d7f72d0989a3567bcdd6c2f34f5f5b0fae41798" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09769", "title": "Bolzano: From Expert-Guided Proof Search to Automated Open-Problem Solving", "published": "2026-10-07T09:52:46Z", "reviews": [ { "id": "2610.09769", "title": "Bolzano: From Expert-Guided Proof Search to Automated Open-Problem Solving", "published": "2026-10-07T09:52:46Z", "decision": "reviewed_deferred", "reason": "已读Bolzano researchloop/4实验：parallelprovers、verifier、summary与显式notes/proofs/output，3800提取openproblem、人工/author验证部分结果。下一批需独立proofvalidation与提取题面准确性，LLMverifier结论不能代专家检查。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Four automated experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 Bolzano’s research loop", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09769v1", "sha256": "8960c90ac2cea0c7ca75dca55bbeea928a215b7472e1be05aa0ccd465ff1b05a", "local_html_sha256": "8960c90ac2cea0c7ca75dca55bbeea928a215b7472e1be05aa0ccd465ff1b05a", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09769v1", "result": "retrieved", "sha256": "8960c90ac2cea0c7ca75dca55bbeea928a215b7472e1be05aa0ccd465ff1b05a" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09763", "title": "Beyond Policy Support: Interaction Constrained Offline Reinforcement Learning for Autonomous Driving", "published": "2026-10-07T09:48:53Z", "reviews": [ { "id": "2610.09763", "title": "Beyond Policy Support: Interaction Constrained Offline Reinforcement Learning for Autonomous Driving", "published": "2026-10-07T09:48:53Z", "decision": "rejected_or_out_of_scope", "reason": "已读ICDP method/experiment：offline actorcritic diffusion motionplanner、contrastive interactiondensityratio regularization，nuPlan/InterPlan和truckcase。是自动驾驶RL非LLMagent，当前范围外，不能据论文运行现实车辆。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experimental Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09763v1", "sha256": "9256c3acc972639fb38172c6c659c8155a5014faa1dcf0cf926132bbfd4e741d", "local_html_sha256": "9256c3acc972639fb38172c6c659c8155a5014faa1dcf0cf926132bbfd4e741d", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09763v1", "result": "retrieved", "sha256": "9256c3acc972639fb38172c6c659c8155a5014faa1dcf0cf926132bbfd4e741d" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.09761", "title": "PARC-Loc: Text-to-Point-Cloud Localization with Partial Assignment and Relational Consistency", "published": "2026-10-07T09:46:34Z", "reviews": [ { "id": "2610.09761", "title": "PARC-Loc: Text-to-Point-Cloud Localization with Partial Assignment and Relational Consistency", "published": "2026-10-07T09:46:34Z", "decision": "reviewed_deferred", "reason": "PARCpartialassignment+pairwise spatial关系和neighborboundary扩充，KITTI360/CityLoc；真实text/pointcloud特征与coarse/fine权重待实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "4 Method", "location": ".cache/oct12-scan/fulltext/2610.09761.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09761.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 206, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09761v1", "sha256": "46f1a5c305eb1b95f11233fc90572b60d78bedf01f021086c6e67204d412295d", "local_path": ".cache/oct12-scan/fulltext/2610.09761.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09761v1", "result": "retrieved", "sha256": "46f1a5c305eb1b95f11233fc90572b60d78bedf01f021086c6e67204d412295d" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09761", "sha256": "2b6f954f1e441ce6232fce252733a4f1e5b87b601575d92b709104e2869b1c3c", "local_path": ".cache/oct12-scan/fulltext/2610.09761.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 09:46:34 UTC (2,312 KB)", "published": "2026-10-07T09:46:34Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09746", "title": "Flow-of-Thought: A Framework for Visual Reasoning", "published": "2026-10-07T09:37:57Z", "reviews": [ { "id": "2610.09746", "title": "Flow-of-Thought: A Framework for Visual Reasoning", "published": "2026-10-07T09:37:57Z", "decision": "reviewed_deferred", "reason": "FoT学SO2/shortestpath轨迹flow后冻结，hypothesis重建energy判断；lockedsynthetic及BLINKtransfer133validation，非通用LLM推理增益。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "4 The Flow-of-Thought Framework", "location": ".cache/oct12-scan/fulltext/2610.09746.json sections text prefix", "characters_read": [ 0, 650 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5 Data and Experiments", "location": ".cache/oct12-scan/fulltext/2610.09746.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 207, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09746v1", "sha256": "720d835db09c91e8a0132e656b4f4a87a11d404fdd6036a1cc984a41abc60854", "local_path": ".cache/oct12-scan/fulltext/2610.09746.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09746v1", "result": "retrieved", "sha256": "720d835db09c91e8a0132e656b4f4a87a11d404fdd6036a1cc984a41abc60854" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09746", "sha256": "b0bac5dbec0a51fe7bce81434e982815903c2a239a92b0115a4ed5913dbca0f9", "local_path": ".cache/oct12-scan/fulltext/2610.09746.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 09:37:57 UTC (324 KB)", "published": "2026-10-07T09:37:57Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09742", "title": "SoccerNet-FoulRet: Retrieving Semantically Similar Soccer Foul Videos", "published": "2026-10-07T09:31:53Z", "reviews": [ { "id": "2610.09742", "title": "SoccerNet-FoulRet: Retrieving Semantically Similar Soccer Foul Videos", "published": "2026-10-07T09:31:53Z", "decision": "reviewed_deferred", "reason": "SoccerFoulRet693humanverifiedprecedent与category两个relevance endpoint；稀疏judgment无credit不等不相关，真实videoembedding/candidategallery待核。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "3 The SoccerNet-FoulRet Benchmark", "location": ".cache/oct12-scan/fulltext/2610.09742.json sections text prefix", "characters_read": [ 0, 850 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "4 Experiments", "section_index": 3, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 208, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09742v1", "sha256": "893cf0f85c878fe4c94de91515ba5b204c932df87a6603570aca389c91a68573", "local_path": ".cache/oct12-scan/fulltext/2610.09742.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09742v1", "result": "retrieved", "sha256": "893cf0f85c878fe4c94de91515ba5b204c932df87a6603570aca389c91a68573" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09742", "sha256": "8d6386861995c9ce984a0d563644bbe96ffa769d8932bed4fd5a87cb55b0e78e", "local_path": ".cache/oct12-scan/fulltext/2610.09742.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 09:31:53 UTC (8,628 KB)", "published": "2026-10-07T09:31:53Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09718", "title": "YUBI-STAG: Contact and Semantic-Rich Alignment for VLAs via Automated Video-Language Grounding", "published": "2026-10-07T09:14:26Z", "reviews": [ { "id": "2610.09718", "title": "YUBI-STAG: Contact and Semantic-Rich Alignment for VLAs via Automated Video-Language Grounding", "published": "2026-10-07T09:14:26Z", "decision": "reviewed_deferred", "reason": "YUBISTAG contact/spatiotemporal语义annotation再蒸馏YUBIVLM并VLAposttrain；bimanual语言/接触闭环需真实video与robot，不用标签fixture。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "6. Experiments", "location": ".cache/oct12-scan/fulltext/2610.09718.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3. Spatio-Temporal Interaction Annotation", "section_index": 2, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 209, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09718v1", "sha256": "38b8065193fd85ab3638696a285f7f4ee6e90a0d4f0ef608a4cd5f03c9d5f516", "local_path": ".cache/oct12-scan/fulltext/2610.09718.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09718v1", "result": "retrieved", "sha256": "38b8065193fd85ab3638696a285f7f4ee6e90a0d4f0ef608a4cd5f03c9d5f516" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09718", "sha256": "3d8a0296b21ba1799887a81df200b15fd65d31b66cb45e73ba971d01c0ade1ad", "local_path": ".cache/oct12-scan/fulltext/2610.09718.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 09:14:26 UTC (6,769 KB)", "published": "2026-10-07T09:14:26Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09710", "title": "SpikingVLA: Asynchronous Spiking Vision-Language-Action Models", "published": "2026-10-07T09:05:07Z", "reviews": [ { "id": "2610.09710", "title": "SpikingVLA: Asynchronous Spiking Vision-Language-Action Models", "published": "2026-10-07T09:05:07Z", "decision": "reviewed_deferred", "reason": "SpikingVLA DIF dendriticmix/adaptivefiring与跨component异步ANN→SNN，三VLNCE；TTFA/total及navigation实测不可拿spike标签替代。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "5 Methods", "location": ".cache/oct12-scan/fulltext/2610.09710.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "6 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09710.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 210, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09710v1", "sha256": "9828c420401d03208afb8f7c3c2b169bec237aee8532960e97b0c5ee0ffe52b1", "local_path": ".cache/oct12-scan/fulltext/2610.09710.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09710v1", "result": "retrieved", "sha256": "9828c420401d03208afb8f7c3c2b169bec237aee8532960e97b0c5ee0ffe52b1" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09710", "sha256": "a6488508a4adfac7518b71c5cc2200aad81a0f562022ca21d67d205caab59bc0", "local_path": ".cache/oct12-scan/fulltext/2610.09710.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 09:05:07 UTC (12,941 KB)", "published": "2026-10-07T09:05:07Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09703", "title": "Understanding and Mitigating Token-Pruning-Induced Vulnerabilities in VLMs", "published": "2026-10-07T09:01:09Z", "reviews": [ { "id": "2610.09703", "title": "Understanding and Mitigating Token-Pruning-Induced Vulnerabilities in VLMs", "published": "2026-10-07T09:01:09Z", "decision": "reviewed_deferred", "reason": "SAP maliciousanchor定位、benigntokenrestore与attentionreallocate，LLaVA固定pruning安全/utility；预计算safetyvector和attackeraware评测需真实hook。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "6 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09703.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "4 Mitigating Token-Pruning-Induced Vulnerabilities", "section_index": 3, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 211, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09703v1", "sha256": "3dfc28d3ddc6332dee74fd3d7e8e4da615103c597ace853e3795ca1f65ba8fa0", "local_path": ".cache/oct12-scan/fulltext/2610.09703.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09703v1", "result": "retrieved", "sha256": "3dfc28d3ddc6332dee74fd3d7e8e4da615103c597ace853e3795ca1f65ba8fa0" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09703", "sha256": "ae2ecbc26a578b26c8fa8897c372a0c7fd1dd480c0d91723b9155a2449230ba6", "local_path": ".cache/oct12-scan/fulltext/2610.09703.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 09:01:09 UTC (10,177 KB)", "published": "2026-10-07T09:01:09Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09700", "title": "What Makes Synthetic Hard Negatives Work in Vision-Language Pretraining?", "published": "2026-10-07T08:58:57Z", "reviews": [ { "id": "2610.09700", "title": "What Makes Synthetic Hard Negatives Work in Vision-Language Pretraining?", "published": "2026-10-07T08:58:57Z", "decision": "reviewed_deferred", "reason": "SNAP排除matchedpositive的intra-modalhardnegative，fixedtemperature避免saturation；CC3M/12M linkrot与matchedCLIP/FLIP数据预算须保留。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "5 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09700.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "4 Adapting Synthetic Negatives to Vision-Language Pretraining", "section_index": 3, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 212, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09700v1", "sha256": "7e2e61a6f0d071a9847d17f6c563994d68c677a3b5d5484ff2b6cd2b378c2ae2", "local_path": ".cache/oct12-scan/fulltext/2610.09700.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09700v1", "result": "retrieved", "sha256": "7e2e61a6f0d071a9847d17f6c563994d68c677a3b5d5484ff2b6cd2b378c2ae2" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09700", "sha256": "a6fa946a0a32c1ce4d69e407213dfd6d1f4c44c1e8f91c502c8155fcc9656ca9", "local_path": ".cache/oct12-scan/fulltext/2610.09700.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 08:58:57 UTC (1,498 KB)", "published": "2026-10-07T08:58:57Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10623", "title": "Recurrent Self-Improvement: Dynamic Cross-Loop On-Policy Distillation for Looped Language Models", "published": "2026-10-07T08:54:30Z", "reviews": [ { "id": "2610.10623", "title": "Recurrent Self-Improvement: Dynamic Cross-Loop On-Policy Distillation for Looped Language Models", "published": "2026-10-07T08:54:30Z", "decision": "reviewed_deferred", "reason": "LoopOPDterminalcomputeloop frozen teacher指导intermediateloop student trajectories，dynamic版按sharedparameters刷新teacher；OuroThinking DAPO与general/code评测，下一批必须真loopbackbone/dynamicweights而非两个不同名普通LM。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "4 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10623v1", "sha256": "829d72fd3aaf1e4f2db3e733b2715d91418a0086979a1f34d669f56050c7a870", "local_html_sha256": "829d72fd3aaf1e4f2db3e733b2715d91418a0086979a1f34d669f56050c7a870", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10623v1", "result": "retrieved", "sha256": "829d72fd3aaf1e4f2db3e733b2715d91418a0086979a1f34d669f56050c7a870" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10622", "title": "WorldBench: Evaluating LLMs on Three.js Voxel World Generation", "published": "2026-10-07T08:52:08Z", "reviews": [ { "id": "2610.10622", "title": "WorldBench: Evaluating LLMs on Three.js Voxel World Generation", "published": "2026-10-07T08:52:08Z", "decision": "reviewed_deferred", "reason": "Three.js world评审同时探索运行渲染/时钟/nav和读源code；codequote需exactsource，视觉可测性质pixel验证，两个证据halves独立计分。; 单固定prompt五frontierworlds，365items双通道差异及人工构造feature-removal mutation测试。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.10622v1#S3", "https://arxiv.org/html/2610.10622v1#S4", "https://arxiv.org/html/2610.10622v1#S5", "https://arxiv.org/html/2610.10622v1#S6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10622v1", "sha256": "1265b42b975f16534213ffcbbd1ce4763a7fcb60c20fe2f12b76c0bc1eae99a5" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09684", "title": "From Pareto to Preference: Personalized Test-Time Scaling via Amortized Agentic Policy Discovery", "published": "2026-10-07T08:46:56Z", "reviews": [ { "id": "2610.09684", "title": "From Pareto to Preference: Personalized Test-Time Scaling via Amortized Agentic Policy Discovery", "published": "2026-10-07T08:46:56Z", "decision": "reviewed_deferred", "reason": "PersonTTS executablecontroller discovery按accuracy/latency/costjoint满足度，requirement-matched经验reuse与frozenGuide，AIME/HMMT source/targetprofiles和heldoutproblems；下一批controller须actualexecution/evaluation，replayrecords不能泄漏test答案。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09684v1", "sha256": "a03218b24a19a33b6a272be8fd51b3c11e010361e13acd3d991a86ca46c42f47", "local_html_sha256": "a03218b24a19a33b6a272be8fd51b3c11e010361e13acd3d991a86ca46c42f47", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09684v1", "result": "retrieved", "sha256": "a03218b24a19a33b6a272be8fd51b3c11e010361e13acd3d991a86ca46c42f47" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09683", "title": "System Switch: When Should a Fast Decision Model Stop and Think?", "published": "2026-10-07T08:44:47Z", "reviews": [ { "id": "2610.09683", "title": "System Switch: When Should a Fast Decision Model Stop and Think?", "published": "2026-10-07T08:44:47Z", "decision": "reviewed_deferred", "reason": "已读SystemSwitch gate/results：实际fast typeddecision与reasoningVLM Doomcontinuousloop，accuracy/calibration/AUROC分开、33game零exit负结果。gate percentile文中取testset，下一批需重设validation-only选择；offline deferral收益不能宣称closedloop能力。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 The System Switch", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09683v1", "sha256": "b1785d50dcae2e6829458555c888a768a99ce73a11d2e7f2cb6bdb179f24484a", "local_html_sha256": "b1785d50dcae2e6829458555c888a768a99ce73a11d2e7f2cb6bdb179f24484a", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09683v1", "result": "retrieved", "sha256": "b1785d50dcae2e6829458555c888a768a99ce73a11d2e7f2cb6bdb179f24484a" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09681", "title": "Cost-Efficient Theorem Proving via Agent Orchestration in Program Verification", "published": "2026-10-07T08:42:24Z", "reviews": [ { "id": "2610.09681", "title": "Cost-Efficient Theorem Proving via Agent Orchestration in Program Verification", "published": "2026-10-07T08:42:24Z", "decision": "reviewed_deferred", "reason": "已读CoCo-Prover method/evaluation：AND/OR hypergraph+lemma DAG，symbolic goaltopologicalselect、priced bounded specialist calls的evolvingrouter，5Leanbench与budgetfrontier。下一批真实Leankernel及所有failurecost、验证集router，不用goldproofscore选择action。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Evaluation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09681v1", "sha256": "ad693ca4aa1b7542f3100224d8a940d1982f1f4828de00da4a3adbeddb3736e8", "local_html_sha256": "ad693ca4aa1b7542f3100224d8a940d1982f1f4828de00da4a3adbeddb3736e8", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09681v1", "result": "retrieved", "sha256": "ad693ca4aa1b7542f3100224d8a940d1982f1f4828de00da4a3adbeddb3736e8" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09665", "title": "SAPD: Step-Aligned Privileged Distillation", "published": "2026-10-07T08:31:37Z", "reviews": [ { "id": "2610.09665", "title": "SAPD: Step-Aligned Privileged Distillation", "published": "2026-10-07T08:31:37Z", "decision": "reviewed_deferred", "reason": "SAPD rolloutfree固定demonstrations按reasoningstep对privileged信息提供context-dependent distributiontargets，Qwen3-4B math/OODeval和loop-speed对照；下一批真实offline teacherdistribution，gold仅traincontext允许。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Step-Aligned Privileged Distillation", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09665v1", "sha256": "9ee60d653a0c3c142bc45011fb59982ba7eca7312172b04251cc0d0633f718df", "local_html_sha256": "9ee60d653a0c3c142bc45011fb59982ba7eca7312172b04251cc0d0633f718df", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09665v1", "result": "retrieved", "sha256": "9ee60d653a0c3c142bc45011fb59982ba7eca7312172b04251cc0d0633f718df" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09647", "title": "When Rank Rises as LLMs Degrade", "published": "2026-10-07T08:22:30Z", "reviews": [ { "id": "2610.09647", "title": "When Rank Rises as LLMs Degrade", "published": "2026-10-07T08:22:30Z", "decision": "reviewed_deferred", "reason": "RankMe/effective-rank在Qwen3-0.6B degradation controls和leave-one-seed-out calibration/test下双向诊断；没早于heldout loss且健康seed false alarm，适合training-health诊断下一批，不能宣传早预警或能力算法。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Setup", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "6 A two-sided monitor under held-out false-alarm evaluation", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09647v1", "sha256": "f1d0015f7de8cdb6c3d64b99680a0e78572b559f609df58d4d975fb628581267", "local_html_sha256": "f1d0015f7de8cdb6c3d64b99680a0e78572b559f609df58d4d975fb628581267", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09647v1", "result": "retrieved", "sha256": "f1d0015f7de8cdb6c3d64b99680a0e78572b559f609df58d4d975fb628581267" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09643", "title": "CircuitATLAS: Agentic reasoning over a systems neuroscience knowledge graph for target discovery in circuitopathies", "published": "2026-10-07T08:19:27Z", "reviews": [ { "id": "2610.09643", "title": "CircuitATLAS: Agentic reasoning over a systems neuroscience knowledge graph for target discovery in circuitopathies", "published": "2026-10-07T08:19:27Z", "decision": "reviewed_deferred", "reason": "已读CircuitATLAS KG/labloop高层：provenance关系和agenthypothesis、human-governed in-vivo validation，巨大private测量/伦理协议成本。科学agent候选保留但本批不可复現湿实验，不运行生物干预设计；真实labcase非通用自主药物发现能力。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2  Structuring systems neuroscience into a knowledge graph using LLMs", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5  Closing the loop from target nomination to experimental validation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09643v1", "sha256": "029ea7b0773f0a5dce7ff2a4c2842f1edcba875bb01964ce03e5f015e2124b29", "local_html_sha256": "029ea7b0773f0a5dce7ff2a4c2842f1edcba875bb01964ce03e5f015e2124b29", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09643v1", "result": "retrieved", "sha256": "029ea7b0773f0a5dce7ff2a4c2842f1edcba875bb01964ce03e5f015e2124b29" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09639", "title": "On-Policy Distillation Teaches New Skills but Not New Knowledge", "published": "2026-10-07T08:15:10Z", "reviews": [ { "id": "2610.09639", "title": "On-Policy Distillation Teaches New Skills but Not New Knowledge", "published": "2026-10-07T08:15:10Z", "decision": "reviewed_deferred", "reason": "控制randomfacts/compositionalstructures独立teacher知识/技能，reverse/forwardKL和studentrolloutfactorial分离OPDtransfer；结论限制四model实验，下一批diagnostic synthetic训练不能goldtable进evaluatedrollout。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Experimental Setting: Separating Knowledge and Skill", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Synthetic Experiment Results", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "5 Real-World Validation", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09639v1", "sha256": "80a3d7103e7a5f4bd6d303213468cd412b47385f08cbd0181a338489a15e023d", "local_html_sha256": "80a3d7103e7a5f4bd6d303213468cd412b47385f08cbd0181a338489a15e023d", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09639v1", "result": "retrieved", "sha256": "80a3d7103e7a5f4bd6d303213468cd412b47385f08cbd0181a338489a15e023d" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09633", "title": "Coding-Agent Benchmarks Should Match Their Users' Task Flows", "published": "2026-10-07T08:08:39Z", "reviews": [ { "id": "2610.09633", "title": "Coding-Agent Benchmarks Should Match Their Users' Task Flows", "published": "2026-10-07T08:08:39Z", "decision": "reviewed_deferred", "reason": "已读SWE-TaskFlow productionrecords/transform方法：4782真实IDE session只分析user方、公开corpora分布不同，calibratedprompt splitting+verifiableQA保原test。700taskpilot多步cost约双而resolve无稳定变动；下一批评测infra，生产使用日志非生产A/B。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 Task Flows of Production Sessions", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 SWE-TaskFlow: Transforming Verified Tasks", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09633v1", "sha256": "bf7a5dd1c76200e974f5e6429c00ce1e36dd9775f402e0b7fb3a7ca03e9292c2", "local_html_sha256": "bf7a5dd1c76200e974f5e6429c00ce1e36dd9775f402e0b7fb3a7ca03e9292c2", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09633v1", "result": "retrieved", "sha256": "bf7a5dd1c76200e974f5e6429c00ce1e36dd9775f402e0b7fb3a7ca03e9292c2" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09624", "title": "How Do Agentic LLMs Decide to Call Tools? A Tool-Call Vector Shaped by Suppression", "published": "2026-10-07T08:01:39Z", "reviews": [ { "id": "2610.09624", "title": "How Do Agentic LLMs Decide to Call Tools? A Tool-Call Vector Shaped by Suppression", "published": "2026-10-07T08:01:39Z", "decision": "reviewed_deferred", "reason": "已读toolcall causalvector分析/跨model评测：500minimalverbpairs300discover200heldout，hiddenstateintervention/transcoder，7families与native tau2 generalization。下一批须真实activations causalnecessity/sufficiency，不用文字heuristic决定toolcall或按test筛vector。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "7 The Tool-Call Vector Across Scales and Model Families", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 One Verb Flips the Call-or-No-Call Decision", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09624v1", "sha256": "3c1e05f7c27315e43a42d8da267b69f853591a33b8e023766cd725e50de85f83", "local_html_sha256": "3c1e05f7c27315e43a42d8da267b69f853591a33b8e023766cd725e50de85f83", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09624v1", "result": "retrieved", "sha256": "3c1e05f7c27315e43a42d8da267b69f853591a33b8e023766cd725e50de85f83" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10619", "title": "TestJack: Should you trust the results in coding benchmarks? Agentic Coding Benchmarks Auditing via Evaluator Evolution", "published": "2026-10-07T07:58:04Z", "reviews": [ { "id": "2610.10619", "title": "TestJack: Should you trust the results in coding benchmarks? Agentic Coding Benchmarks Auditing via Evaluator Evolution", "published": "2026-10-07T07:58:04Z", "decision": "reviewed_deferred", "reason": "已读TestJack method/evaluation：pertrial要求生成test，仅外部groundtruthpatch通过的test保留、replay确认fail，randomaudit共享降低cost。下一批evaluatorinfra必须oracle与被评agent隔离，多validimplementation公平性待审，不把fixedtests通过当speccomplete。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Evaluation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10619v1", "sha256": "887e7a852b67bef025f8347f783500406a3243312a9893d7976b89230467558a", "local_html_sha256": "887e7a852b67bef025f8347f783500406a3243312a9893d7976b89230467558a", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10619v1", "result": "retrieved", "sha256": "887e7a852b67bef025f8347f783500406a3243312a9893d7976b89230467558a" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09597", "title": "COPC: Coupled Off-Policy Correction for Asynchronous LLM Reinforcement Learning", "published": "2026-10-07T07:43:49Z", "reviews": [ { "id": "2610.09597", "title": "COPC: Coupled Off-Policy Correction for Asynchronous LLM Reinforcement Learning", "published": "2026-10-07T07:43:49Z", "decision": "reviewed_deferred", "reason": "COPC协调tokenratiomasking与TDresidual two-sidedclippedreturns，policy/advantagestaleness耦合，math/search异步64stepstale与synchronousspeed对照；下一批需真实actorcritic/stale rollout和jointval参数选择。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "5 COPC: Coupled Off-Policy Correction", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 1300 ] }, { "heading": "6 experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09597v1", "sha256": "f3db33889b29fb0954fa0bf94e17eab1943fc7ea5caf580217ae73ac7a1c6a48", "local_html_sha256": "f3db33889b29fb0954fa0bf94e17eab1943fc7ea5caf580217ae73ac7a1c6a48", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09597v1", "result": "retrieved", "sha256": "f3db33889b29fb0954fa0bf94e17eab1943fc7ea5caf580217ae73ac7a1c6a48" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09590", "title": "Learning Situation-Conditioned Thinking Policies for Long-Term LLM Agents", "published": "2026-10-07T07:34:08Z", "reviews": [ { "id": "2610.09590", "title": "Learning Situation-Conditioned Thinking Policies for Long-Term LLM Agents", "published": "2026-10-07T07:34:08Z", "decision": "reviewed_deferred", "reason": "已读thinkingmemory method/verification：lightweight situationpolicy加cross-episode发现、模拟robot异构signals controlledrules，perfectF1/微秒统计是该机制testbed。下一批保留L1诊断边界，不能将合成规律推广长程LLM自主学习。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "III Situation-Conditioned Thinking Memory with Cross-Experience Knowledge Discovery", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "IV Verification", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09590v1", "sha256": "d5a885dbfe4d11fba996d01f00b78eddc72f87558af493a6f1fce0084e340997", "local_html_sha256": "d5a885dbfe4d11fba996d01f00b78eddc72f87558af493a6f1fce0084e340997", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09590v1", "result": "retrieved", "sha256": "d5a885dbfe4d11fba996d01f00b78eddc72f87558af493a6f1fce0084e340997" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09587", "title": "Collaborative Reasoning Distillation via Cross-Feedback and Coherent Curation", "published": "2026-10-07T07:31:33Z", "reviews": [ { "id": "2610.09587", "title": "Collaborative Reasoning Distillation via Cross-Feedback and Coherent Curation", "published": "2026-10-07T07:31:33Z", "decision": "reviewed_deferred", "reason": "CRD multi-teachercrosscritique、stepqualityjudge、coherentstitch/RQO预算curriculum，50k math/code/logic/STEM及contamination检查；下一批需真实judge/教师生成与coherence操作，保留benchmarkclaims可验证边界。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "2 Methodology", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Evaluation", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09587v1", "sha256": "a897e8e1a5618bdf7ff1d97f9b45dcc6f1626b5e07954624127cdad58f910b74", "local_html_sha256": "a897e8e1a5618bdf7ff1d97f9b45dcc6f1626b5e07954624127cdad58f910b74", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09587v1", "result": "retrieved", "sha256": "a897e8e1a5618bdf7ff1d97f9b45dcc6f1626b5e07954624127cdad58f910b74" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10616", "title": "When Routing Reveals Membership: Privacy Leakage from MoE Router Telemetry", "published": "2026-10-07T07:26:51Z", "reviews": [ { "id": "2610.10616", "title": "When Routing Reveals Membership: Privacy Leakage from MoE Router Telemetry", "published": "2026-10-07T07:26:51Z", "decision": "reviewed_deferred", "reason": "shadowMoE训练membershipclassifier用router统计+output ensemble，三架构三域；privacy评测需shadow/targetdata严格分离，受限telemetry保真tradeoff。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "5 Experiments", "location": ".cache/oct12-scan/fulltext/2610.10616.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "4 Router-Augmented Membership Inference", "section_index": 3, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 213, "primary_fulltext": { "url": "https://arxiv.org/html/2610.10616v1", "sha256": "ce5794a89165439d6e53820f1a1609efde68293b1217d773f7f46c3c3f58e6b6", "local_path": ".cache/oct12-scan/fulltext/2610.10616.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10616v1", "result": "retrieved", "sha256": "ce5794a89165439d6e53820f1a1609efde68293b1217d773f7f46c3c3f58e6b6" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.10616", "sha256": "9754731b154689512dd641f95648a0aff2975722f5316fc2383a820f3dcd616b", "local_path": ".cache/oct12-scan/fulltext/2610.10616.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 07:26:51 UTC (1,117 KB)", "published": "2026-10-07T07:26:51Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09581", "title": "Correct Answers, Unsupported Findings: Evidence Binding in Forensic Reconstruction of LLM Agent Logs", "published": "2026-10-07T07:26:33Z", "reviews": [ { "id": "2610.09581", "title": "Correct Answers, Unsupported Findings: Evidence Binding in Forensic Reconstruction of LLM Agent Logs", "published": "2026-10-07T07:26:33Z", "decision": "reviewed_deferred", "reason": "已读forensic reconstruction method/results：64AgentDojo savedcases，literalagreement、sourcebindings、groundedfindings与abstention分离，deterministicsamepacket有明确boundedtask。下一批auditinfra需保存ID→source绑定，答对source不等有证据、guess不能作trace因果。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "Reconstruction method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09581v1", "sha256": "1caffcf26bd6b60aeedc1c878cf1890f19857d80124e66fee3dd8e0a01a5078c", "local_html_sha256": "1caffcf26bd6b60aeedc1c878cf1890f19857d80124e66fee3dd8e0a01a5078c", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09581v1", "result": "retrieved", "sha256": "1caffcf26bd6b60aeedc1c878cf1890f19857d80124e66fee3dd8e0a01a5078c" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10612", "title": "PyCache Trap: The Inspection-Execution Gap in Agent Skill Scanners", "published": "2026-10-07T07:10:11Z", "reviews": [ { "id": "2610.10612", "title": "PyCache Trap: The Inspection-Execution Gap in Agent Skill Scanners", "published": "2026-10-07T07:10:11Z", "decision": "reviewed_deferred", "reason": "已读PyCacheTrap method/experiment高层：sourceinspection与Pythoncache实际选取不一致，EAV trustedcompiledartifact reproduction+typedexecutiongraph；scanneradmission、semanticrecognition、runtime分开。下一批安全infra须bytehash/loader固定与sandbox，不执行可疑cache于宿主。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10612v1", "sha256": "a1dd54ed11067c5cafe5cc48cccbe2cc1d3a335b5ac68ac93c1842c0229a3d43", "local_html_sha256": "a1dd54ed11067c5cafe5cc48cccbe2cc1d3a335b5ac68ac93c1842c0229a3d43", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10612v1", "result": "retrieved", "sha256": "a1dd54ed11067c5cafe5cc48cccbe2cc1d3a335b5ac68ac93c1842c0229a3d43" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09563", "title": "EvoSignal: LLM-Guided Evolutionary Design of Modular Traffic Signal Control Programs", "published": "2026-10-07T07:04:22Z", "reviews": [ { "id": "2610.09563", "title": "EvoSignal: LLM-Guided Evolutionary Design of Modular Traffic Signal Control Programs", "published": "2026-10-07T07:04:22Z", "decision": "reviewed_deferred", "reason": "EvoSignalLLM演化modulartrafficcontrol程序，以拥堵反馈保存多objective tradeoffs并zeroshotnetworktransfer；agent-programsearch下一批需实际CityFlowsimulator/程序执行与heldoutflow，不能只有registry标签。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09563v1", "sha256": "7b1ee8ae5f784aef00d1e3ded614c2ac4ddd5a181b4d6158c3815023f07f3439", "local_html_sha256": "7b1ee8ae5f784aef00d1e3ded614c2ac4ddd5a181b4d6158c3815023f07f3439", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09563v1", "result": "retrieved", "sha256": "7b1ee8ae5f784aef00d1e3ded614c2ac4ddd5a181b4d6158c3815023f07f3439" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09550", "title": "Visual Evidence Under Cross-Examination: Evaluating and Controlling Decision-Level Evidence Use in Vision-Language Models", "published": "2026-10-07T06:48:44Z", "reviews": [ { "id": "2610.09550", "title": "Visual Evidence Under Cross-Examination: Evaluating and Controlling Decision-Level Evidence Use in Vision-Language Models", "published": "2026-10-07T06:48:44Z", "decision": "reviewed_deferred", "reason": "CROSS/RIVET preservebinding、candidateconditioncompose与strengthbound，invalidation/rebinding测effectdestination；真实decisionlayer/rawvisualevidence需核。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "6 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09550.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "5 RIVET: A Candidate-Bound Evidence-to-Decision Interface", "section_index": 4, "characters_read": [ 0, 1418 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 214, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09550v1", "sha256": "69d738854482214f8cd9bcbffcffb82324b44e12e16dc8b4255fb5253b1f9fbd", "local_path": ".cache/oct12-scan/fulltext/2610.09550.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09550v1", "result": "retrieved", "sha256": "69d738854482214f8cd9bcbffcffb82324b44e12e16dc8b4255fb5253b1f9fbd" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09550", "sha256": "fd1ff66dc950b82f8bcaeb745be6aefd589b095cf01ceebe6dd1144b4d7ff7d1", "local_path": ".cache/oct12-scan/fulltext/2610.09550.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 06:48:44 UTC (3,930 KB)", "published": "2026-10-07T06:48:44Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.10610", "title": "Code Understanding is a Bottleneck for Coding Agents", "published": "2026-10-07T06:29:42Z", "reviews": [ { "id": "2610.10610", "title": "Code Understanding is a Bottleneck for Coding Agents", "published": "2026-10-07T06:29:42Z", "decision": "reviewed_deferred", "reason": "已读CABRA生成/experimentalsetup：callgraphtransformation控制traversal/search/runtimeresolution/instructionaxis、6840synthetictask及SWEtrace关联。下一批明确controlleddiagnostic，不以grep可解toy任务证明通用codingcapability。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 Cabra : Taking Control of Coding Agent Evaluations", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10610v1", "sha256": "8dc050867f41e45819fb52d46ce5c1c484099df0e53495e322db578f77fb41b7", "local_html_sha256": "8dc050867f41e45819fb52d46ce5c1c484099df0e53495e322db578f77fb41b7", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10610v1", "result": "retrieved", "sha256": "8dc050867f41e45819fb52d46ce5c1c484099df0e53495e322db578f77fb41b7" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09517", "title": "It Is Not Seeing the Hazard: A Frozen Vision-Language Safety Score Measures Its Caption Bank", "published": "2026-10-07T06:15:23Z", "reviews": [ { "id": "2610.09517", "title": "It Is Not Seeing the Hazard: A Frozen Vision-Language Safety Score Measures Its Caption Bank", "published": "2026-10-07T06:15:23Z", "decision": "reviewed_deferred", "reason": "冻结CLIP prompt-margin safety passive scorer，用VLM-free trajectories、geometry-matched pre-contact controls分离caption/viewpoint混杂；下一批evaluation integrity候选，不以policy returns当hazard perception证据。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Measurement protocol", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Results", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09517v1", "sha256": "dc8be4442badc95545d08301558b75b8750b01b944d83ed822185fff9a1c4fe6", "local_html_sha256": "dc8be4442badc95545d08301558b75b8750b01b944d83ed822185fff9a1c4fe6", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09517v1", "result": "retrieved", "sha256": "dc8be4442badc95545d08301558b75b8750b01b944d83ed822185fff9a1c4fe6" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09514", "title": "STRIKE: Learning Visual State Transitions for Physical World Modeling", "published": "2026-10-07T06:11:17Z", "reviews": [ { "id": "2610.09514", "title": "STRIKE: Learning Visual State Transitions for Physical World Modeling", "published": "2026-10-07T06:11:17Z", "decision": "reviewed_deferred", "reason": "STRIKE observedtrainingstates+timedtransition生成与stateconditiondynamics，PhysicalIQ/PhyGen/RoboTwin；prospectiveplanner不可读未来监督。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "4. Method", "location": ".cache/oct12-scan/fulltext/2610.09514.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5. Experiments", "location": ".cache/oct12-scan/fulltext/2610.09514.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 215, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09514v1", "sha256": "089662bc02cd11c33aa9068a937d2ed2797f37749bf9bebbca97d9107aea5a3b", "local_path": ".cache/oct12-scan/fulltext/2610.09514.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09514v1", "result": "retrieved", "sha256": "089662bc02cd11c33aa9068a937d2ed2797f37749bf9bebbca97d9107aea5a3b" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09514", "sha256": "28579aaa90cf7531f4344b977a422b8ce65755b33d702675b9689ec4ad794bd5", "local_path": ".cache/oct12-scan/fulltext/2610.09514.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 06:11:17 UTC (2,985 KB)", "published": "2026-10-07T06:11:17Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09513", "title": "OmniCam: Omni-Camera Trajectory Generation via Geometry-Grounded Pose Token Learning", "published": "2026-10-07T06:08:28Z", "reviews": [ { "id": "2610.09513", "title": "OmniCam: Omni-Camera Trajectory Generation via Geometry-Grounded Pose Token Learning", "published": "2026-10-07T06:08:28Z", "decision": "reviewed_deferred", "reason": "OmniCam panoramicpointcloud+hybridpose/quaternion+3Dtargetanchor，OmniCaT267k；scene-levelsplitmanifest尚需核，robotdemo非quant部署收益。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "4 Method", "location": ".cache/oct12-scan/fulltext/2610.09513.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09513.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 216, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09513v1", "sha256": "a4d7a68cf24abc4341a31cfb6bb5aebdbd17ffe7fd0cd066f2b8fb25fe495f48", "local_path": ".cache/oct12-scan/fulltext/2610.09513.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09513v1", "result": "retrieved", "sha256": "a4d7a68cf24abc4341a31cfb6bb5aebdbd17ffe7fd0cd066f2b8fb25fe495f48" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09513", "sha256": "0626e1cfaeea50aff1e7e0fc4e8cd9dc2e6f7d46d0cf3959e3c63c0aa1a08041", "local_path": ".cache/oct12-scan/fulltext/2610.09513.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 06:08:28 UTC (32,437 KB)", "published": "2026-10-07T06:08:28Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09496", "title": "Sparse Feature Policy Unlearning Mitigates State Hallucination in Vision-Language-Action Models", "published": "2026-10-07T05:52:02Z", "reviews": [ { "id": "2610.09496", "title": "Sparse Feature Policy Unlearning Mitigates State Hallucination in Vision-Language-Action Models", "published": "2026-10-07T05:52:02Z", "decision": "reviewed_deferred", "reason": "SOUL SAE hallucinationfeatureforget/cleanretain policyunlearning，OpenVLA/pi0.5、LIBEROPlus/RoboCasa/Franka；feature选择仅train/validation。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "V Experiments", "location": ".cache/oct12-scan/fulltext/2610.09496.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "IV S parse Feature P o licy U n l earning ( SOUL )", "section_index": 3, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 217, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09496v1", "sha256": "92d95e090b0a2978c2f762455f7c8b0f2e7b855705c698bb7080f9586fb6d6b0", "local_path": ".cache/oct12-scan/fulltext/2610.09496.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09496v1", "result": "retrieved", "sha256": "92d95e090b0a2978c2f762455f7c8b0f2e7b855705c698bb7080f9586fb6d6b0" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09496", "sha256": "640f29122bda7b8c20cea69e762b7194cc0c72f0405fdc2f5073033187cfcc8e", "local_path": ".cache/oct12-scan/fulltext/2610.09496.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 05:52:02 UTC (10,784 KB)", "published": "2026-10-07T05:52:02Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09489", "title": "Goldsmith: Gold-Loss-Guided Definition Optimization with an Agentic Annotation Harness", "published": "2026-10-07T05:45:25Z", "reviews": [ { "id": "2610.09489", "title": "Goldsmith: Gold-Loss-Guided Definition Optimization with an Agentic Annotation Harness", "published": "2026-10-07T05:45:25Z", "decision": "reviewed_deferred", "reason": "已读Goldsmith method/experiment：小expertcalibrationgold训练textdefinition、structuredloss下降才accept，外harnessschema/retrieval/repair/judge/humanreview固定。训练editor看calibrationgold合法，deploymentannotation需隔离测试label；下一批promptoptimization而非weighttraining。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09489v1", "sha256": "727545c2f5b4a3b0359e0b421feca0d7f24a9e83b96324ad1d535ec1bbde52e7", "local_html_sha256": "727545c2f5b4a3b0359e0b421feca0d7f24a9e83b96324ad1d535ec1bbde52e7", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09489v1", "result": "retrieved", "sha256": "727545c2f5b4a3b0359e0b421feca0d7f24a9e83b96324ad1d535ec1bbde52e7" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09484", "title": "MIMESIS: Learning User Simulators as Training Environments for Interactive Agents", "published": "2026-10-07T05:37:09Z", "reviews": [ { "id": "2610.09484", "title": "MIMESIS: Learning User Simulators as Training Environments for Interactive Agents", "published": "2026-10-07T05:37:09Z", "decision": "reviewed_deferred", "reason": "Meta优先。已读MIMESIS user simulator及agent posttraining：user-side midtrain、ThoughtTrace SFT、多域GRPO后冻结simulator，再agent GRPO+CSD token重加权；private thought仅retrospective teacher训练上下文。9 unseen user评测和8环境，下一批须真实双训练链而非fixed simulator。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 MIMESIS: Learning a Realistic User Simulator", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 11063 ] }, { "heading": "5 Agent Post-Training with a Learned User Simulator", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 8902 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09484v1", "sha256": "a3e31729a16687b8076c1b9cbfdc369cc560f217a3805c27d49cb7df38e35dca", "local_html_sha256": "a3e31729a16687b8076c1b9cbfdc369cc560f217a3805c27d49cb7df38e35dca", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09484v1", "result": "retrieved", "sha256": "a3e31729a16687b8076c1b9cbfdc369cc560f217a3805c27d49cb7df38e35dca" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09478", "title": "InstanceBench: Diagnosing Referential Reasoning and Target Identity in Referring Expression Segmentation", "published": "2026-10-07T05:34:00Z", "reviews": [ { "id": "2610.09478", "title": "InstanceBench: Diagnosing Referential Reasoning and Target Identity in Referring Expression Segmentation", "published": "2026-10-07T05:34:00Z", "decision": "reviewed_deferred", "reason": "InstanceBench TCES相同target多referentialcue、identity-awaremaskmetrics，22checkpoints与训练干预；需真实native mask和target/maskfailure分解。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "3 Evaluation Protocol", "location": ".cache/oct12-scan/fulltext/2610.09478.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "4 Experiments", "section_index": 3, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 218, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09478v1", "sha256": "3333c84c2028010ab8b5be27727428009d9349d30ebe9736e135ad4a97e8934c", "local_path": ".cache/oct12-scan/fulltext/2610.09478.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09478v1", "result": "retrieved", "sha256": "3333c84c2028010ab8b5be27727428009d9349d30ebe9736e135ad4a97e8934c" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09478", "sha256": "17659814dbd3a7870e381d307e29a4d02c7d1b583a08afb270a367082dd67916", "local_path": ".cache/oct12-scan/fulltext/2610.09478.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 05:34:00 UTC (4,011 KB)", "published": "2026-10-07T05:34:00Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09476", "title": "CHASE: Channel-Aligned Structure Exploitation for Geometry-Aware Model Engineering", "published": "2026-10-07T05:33:00Z", "reviews": [ { "id": "2610.09476", "title": "CHASE: Channel-Aligned Structure Exploitation for Geometry-Aware Model Engineering", "published": "2026-10-07T05:33:00Z", "decision": "reviewed_deferred", "reason": "CHASE GSA用于rotationadaptation/pruning/merge及CAGA/SAKV/CAPS；读到COEC公开WikiText结果和CAGA构造，后段详细方法/评测尚需专项，defer不称六项全实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "additional targeted method/evaluation", "heading": "3 CHANNEL-ALIGNED STRUCTURE EXPLOITATION", "section_index": 2, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "3 CHANNEL-ALIGNED STRUCTURE EXPLOITATION", "section_index": 2, "characters_read": [ 5000, 8000 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "3 CHANNEL-ALIGNED STRUCTURE EXPLOITATION", "location": "text[13000:15000] CAGA construction", "characters_read": [ 13000, 15000 ], "scope": "bounded excerpt; raw source available, cached section extraction may be capped" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 219, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09476v1", "sha256": "49118c0c64da5cadaf8e1586db30b69ba451abf23fd5ee4606f4243cb4d0a1a7", "local_path": ".cache/oct12-scan/fulltext/2610.09476.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09476v1", "result": "retrieved", "sha256": "49118c0c64da5cadaf8e1586db30b69ba451abf23fd5ee4606f4243cb4d0a1a7" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09476", "sha256": "3bf3d4938b570e3a8f37c1c542dfbc2399d0a49a89212c27a19c89ee5bf1cccc", "local_path": ".cache/oct12-scan/fulltext/2610.09476.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 05:33:00 UTC (64 KB)", "published": "2026-10-07T05:33:00Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09471", "title": "When Should an In-Context Learner Expand Its Hypothesis Space?", "published": "2026-10-07T05:26:15Z", "reviews": [ { "id": "2610.09471", "title": "When Should an In-Context Learner Expand Its Hypothesis Space?", "published": "2026-10-07T05:26:15Z", "decision": "reviewed_deferred", "reason": "SRE exact Bayesian one-shot structural family expansion价值边界与learner utility训练，cost/horizon独立控制；受控理论诊断符合机制研究但非现成agent capability，下一批需gold-free action policy与明确diagnostic等级。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Structural Revision Environment", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Results", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09471v1", "sha256": "26b2e479679184e5339afb2b5405e08a7290e5a75280b693c5e88005fc3d9957", "local_html_sha256": "26b2e479679184e5339afb2b5405e08a7290e5a75280b693c5e88005fc3d9957", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09471v1", "result": "retrieved", "sha256": "26b2e479679184e5339afb2b5405e08a7290e5a75280b693c5e88005fc3d9957" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09467", "title": "Boundary-Free Contextual Biasing: Depth-Adaptive Gating and Reading-Space Matching for Unsegmented Languages", "published": "2026-10-07T05:23:18Z", "reviews": [ { "id": "2610.09467", "title": "Boundary-Free Contextual Biasing: Depth-Adaptive Gating and Reading-Space Matching for Unsegmented Languages", "published": "2026-10-07T05:23:18Z", "decision": "reviewed_deferred", "reason": "CTC characterAhoCorasick depthgate/readingspacematch，JAmodel-freebenchmark/两publiccheckpoint/中文迁移；JAdepth参数有后验length选择须保留。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "2 Method", "location": ".cache/oct12-scan/fulltext/2610.09467.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "3 Open JA biasing benchmark", "location": ".cache/oct12-scan/fulltext/2610.09467.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "4 Experiment and results", "section_index": 3, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 220, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09467v1", "sha256": "35a4a774036e6e055c38bfb785618e7af6425ba13f3014c944285c21ab4d12ee", "local_path": ".cache/oct12-scan/fulltext/2610.09467.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09467v1", "result": "retrieved", "sha256": "35a4a774036e6e055c38bfb785618e7af6425ba13f3014c944285c21ab4d12ee" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09467", "sha256": "d86a5f70a17e53d6185577addc51f1ed0bf92e2c688fb31ecadafaed9c6708db", "local_path": ".cache/oct12-scan/fulltext/2610.09467.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 05:23:18 UTC (46 KB)", "published": "2026-10-07T05:23:18Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09466", "title": "LLM-Enabled UAV Dispatch: A System-Level Survey and Taxonomy", "published": "2026-10-07T05:22:51Z", "reviews": [ { "id": "2610.09466", "title": "LLM-Enabled UAV Dispatch: A System-Level Survey and Taxonomy", "published": "2026-10-07T05:22:51Z", "decision": "rejected_or_out_of_scope", "reason": "已读UAVdispatch survey/prelim与pipelinetaxonomy：四类systemlevel综述与安全/latency挑战，无独立新可执行算法/实证bench，本批范围外参考，不授权UAV物理操作。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "II Preliminaries", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "III Pipeline Dispatch", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09466v1", "sha256": "dadc56fd8ae1534741b9af0a610143b4b50c9ebb0a83b9358e9a34dc59edfdbe", "local_html_sha256": "dadc56fd8ae1534741b9af0a610143b4b50c9ebb0a83b9358e9a34dc59edfdbe", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09466v1", "result": "retrieved", "sha256": "dadc56fd8ae1534741b9af0a610143b4b50c9ebb0a83b9358e9a34dc59edfdbe" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.09457", "title": "DSReg: Provably Recovering Individual World Latents without Reconstruction", "published": "2026-10-07T05:12:39Z", "reviews": [ { "id": "2610.09457", "title": "DSReg: Provably Recovering Individual World Latents without Reconstruction", "published": "2026-10-07T05:12:39Z", "decision": "reviewed_deferred", "reason": "DSReg structuraldiversity+Jacobiandependencysparsity从linearidentifiable JEPA分离latent；synthetic/visual/render评测，保证依赖假设非任意foundation identifiability。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09457.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 Identifiability from Structural Diversity", "section_index": 2, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 221, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09457v1", "sha256": "cd6d594b40298e5c5c3f7ed5d2fb6b59c984db77d9099c5cb70c04c305dd123f", "local_path": ".cache/oct12-scan/fulltext/2610.09457.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09457v1", "result": "retrieved", "sha256": "cd6d594b40298e5c5c3f7ed5d2fb6b59c984db77d9099c5cb70c04c305dd123f" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09457", "sha256": "5e1d88229274ca9bf720323f487dc25580dd4aee57bfbbb2db6dc7bd7760a815", "local_path": ".cache/oct12-scan/fulltext/2610.09457.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 05:12:39 UTC (4,496 KB)", "published": "2026-10-07T05:12:39Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09440", "title": "Mixture of Layers: Dynamic Layer Routing for Visual Reasoning", "published": "2026-10-07T04:50:48Z", "reviews": [ { "id": "2610.09440", "title": "Mixture of Layers: Dynamic Layer Routing for Visual Reasoning", "published": "2026-10-07T04:50:48Z", "decision": "reviewed_deferred", "reason": "Meta RealityLabs合作MoL textquery引导layer/patchtopk与reservegate，七finevisualbench；表有负增益且BestDelta为variant最大值，需validation选择。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Mixture of Layers : Selective Visual Representations for Fine-Grained Visual Reasoning", "location": ".cache/oct12-scan/fulltext/2610.09440.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09440.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 Mixture of Layers : Selective Visual Representations for Fine-Grained Visual Reasoning", "section_index": 2, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "4 Experiments", "section_index": 3, "characters_read": [ 0, 1800 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "3 Mixture of Layers : Selective Visual Representations for Fine-Grained Visual Reasoning", "section_index": 2, "characters_read": [ 1800, 3600 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 222, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09440v1", "sha256": "3de331145c8014ea43fb7adb460b9ac1994769ac7bc1cf6f51e8675938bb8802", "local_path": ".cache/oct12-scan/fulltext/2610.09440.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09440v1", "result": "retrieved", "sha256": "3de331145c8014ea43fb7adb460b9ac1994769ac7bc1cf6f51e8675938bb8802" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09440", "sha256": "da6b9092f688027a84459b607bc0e42e45b1286b92c736572233dea9cb3fd9d2", "local_path": ".cache/oct12-scan/fulltext/2610.09440.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 04:50:48 UTC (10,653 KB)", "published": "2026-10-07T04:50:48Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09430", "title": "SkillCycle: Co-Evolving Agent Policies and Skill Banks", "published": "2026-10-07T04:33:48Z", "reviews": [ { "id": "2610.09430", "title": "SkillCycle: Co-Evolving Agent Policies and Skill Banks", "published": "2026-10-07T04:33:48Z", "decision": "reviewed_deferred", "reason": "SkillCycle bank/router冻结policy学习后冻结policy修rules，rule/bank环境配对验证；ALFWorld/WebShop真实controller/distill待实现，no-skill部署不可带bank。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "2 Method: SkillCycle", "location": ".cache/oct12-scan/fulltext/2610.09430.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "3 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09430.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 223, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09430v1", "sha256": "c42399134db81ece9ca9c9c9e34ca05398fb7435517972d833c058b6940a2b9b", "local_path": ".cache/oct12-scan/fulltext/2610.09430.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09430v1", "result": "retrieved", "sha256": "c42399134db81ece9ca9c9c9e34ca05398fb7435517972d833c058b6940a2b9b" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09430", "sha256": "10f929c5886fbdcf6bb13ad635fd5991048c8800d4e3f7c8ca599a033bc3517d", "local_path": ".cache/oct12-scan/fulltext/2610.09430.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 04:33:48 UTC (836 KB)", "published": "2026-10-07T04:33:48Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09408", "title": "TiTok: Audio-Visual LLM for Multi-Segment Temporal Grounding", "published": "2026-10-07T04:08:44Z", "reviews": [ { "id": "2610.09408", "title": "TiTok: Audio-Visual LLM for Multi-Segment Temporal Grounding", "published": "2026-10-07T04:08:44Z", "decision": "reviewed_deferred", "reason": "TiTok time-token interleave+decoupledglobal/local/count/precision/format GDPO，UnAV100multiinterval/Charades；真实audio-video/timevocab与CountF1待核。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Method", "location": ".cache/oct12-scan/fulltext/2610.09408.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09408.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 224, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09408v1", "sha256": "490727036af14fc16ff09a47098ae293ee93f6e2a124a0ebd3a49923536d024d", "local_path": ".cache/oct12-scan/fulltext/2610.09408.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09408v1", "result": "retrieved", "sha256": "490727036af14fc16ff09a47098ae293ee93f6e2a124a0ebd3a49923536d024d" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09408", "sha256": "55220bba17c6794e60e44156e7972569c8557ae8ab5cb00fa0cb3f5ee8bcc965", "local_path": ".cache/oct12-scan/fulltext/2610.09408.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 04:08:44 UTC (923 KB)", "published": "2026-10-07T04:08:44Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09400", "title": "TutorLoop: Regulating Student Learning Behaviors via Sensor-in-the-Loop Generative Feedback", "published": "2026-10-07T04:03:32Z", "reviews": [ { "id": "2610.09400", "title": "TutorLoop: Regulating Student Learning Behaviors via Sensor-in-the-Loop Generative Feedback", "published": "2026-10-07T04:03:32Z", "decision": "reviewed_deferred", "reason": "已读TutorLoop system/evaluation：webcam cognitiveestimate→DRL反馈type→LLM呈现、studentdigitaltwin离线训及187human新task。下一批需真实sensormodel/RL链和隐私同意；有限用户研究不推广所有教学场景，不采集用户camera。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3. TutorLoop: System Design", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4. Evaluation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09400v1", "sha256": "17bb4f8a6c4527491054d004afc1ba2c58882faa34f66bd20594ce5adcff4031", "local_html_sha256": "17bb4f8a6c4527491054d004afc1ba2c58882faa34f66bd20594ce5adcff4031", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09400v1", "result": "retrieved", "sha256": "17bb4f8a6c4527491054d004afc1ba2c58882faa34f66bd20594ce5adcff4031" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09392", "title": "Quantifying Volumetric Risk: Class-Aware Asymmetric Weighted Conformal Prediction for 3D Medical Image Segmentation", "published": "2026-10-07T03:51:16Z", "reviews": [ { "id": "2610.09392", "title": "Quantifying Volumetric Risk: Class-Aware Asymmetric Weighted Conformal Prediction for 3D Medical Image Segmentation", "published": "2026-10-07T03:51:16Z", "decision": "reviewed_deferred", "reason": "CA-WCP densityratio directionalvolumequantile/classasymmetry，BraTS和syntheticCT；weightedexchangeability覆盖不是conditionalclinical保证，calibration/test要分离。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Method", "location": ".cache/oct12-scan/fulltext/2610.09392.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09392.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 225, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09392v1", "sha256": "4b5e5ab426ee3b67aa78ba365aaea5d2b940f2534b3f8c8f786b40a95effbe90", "local_path": ".cache/oct12-scan/fulltext/2610.09392.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09392v1", "result": "retrieved", "sha256": "4b5e5ab426ee3b67aa78ba365aaea5d2b940f2534b3f8c8f786b40a95effbe90" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09392", "sha256": "1f5acd6c5314775ab71b39d2930362e225f6a1c2a91d8149db52dd4663138856", "local_path": ".cache/oct12-scan/fulltext/2610.09392.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 03:51:16 UTC (5,929 KB)", "published": "2026-10-07T03:51:16Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09384", "title": "The Persona Hierarchy Model: Understanding Contextual Generalization in Fine-Tuning LLMs", "published": "2026-10-07T03:40:44Z", "reviews": [ { "id": "2610.09384", "title": "The Persona Hierarchy Model: Understanding Contextual Generalization in Fine-Tuning LLMs", "published": "2026-10-07T03:40:44Z", "decision": "reviewed_deferred", "reason": "Persona Hierarchy/PPR在reference persona下约束initial-model行为，120个prefix×behavior模型/heldout评测及SFT/RL应用；Google DeepMind线索为作者当前去向而非本研究官方机构，优先审查但不夸大Google论文署名；下一批PPR需跨prefix真实KL与reward-hacking评测。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "2 Persona Hierarchy Model", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "6 Persona-preserving regularization", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Experimental Setup", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 The narrowness of contextual generalization is correlated with the proximity with default persona", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09384v1", "sha256": "aec40368df6af778ccaafc5b1e31a8d443c369f8aa7ecb6812e0c0066c5084a6", "local_html_sha256": "aec40368df6af778ccaafc5b1e31a8d443c369f8aa7ecb6812e0c0066c5084a6", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09384v1", "result": "retrieved", "sha256": "aec40368df6af778ccaafc5b1e31a8d443c369f8aa7ecb6812e0c0066c5084a6" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09375", "title": "From Retrieval to Customer Context: Evaluating Frontier-Model Systems for Voice-of-Customer Analysis", "published": "2026-10-07T03:30:22Z", "reviews": [ { "id": "2610.09375", "title": "From Retrieval to Customer Context: Evaluating Frontier-Model Systems for Voice-of-Customer Analysis", "published": "2026-10-07T03:30:22Z", "decision": "reviewed_deferred", "reason": "Customercontextgraph同9432Cursorcorpus三agent30query比较；rubric在28/30题校准缺isolatedtest，comparative0.961非部署效果，defer诊断。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "4 Dataset and Benchmark", "location": ".cache/oct12-scan/fulltext/2610.09375.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 Compared Systems", "section_index": 2, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "5 Automatic Evaluation", "section_index": 4, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 226, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09375v1", "sha256": "fdefe8e4854072de0761378c4b1cf3ddf70fe4a7d2d80f40ad7771e3581556ab", "local_path": ".cache/oct12-scan/fulltext/2610.09375.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09375v1", "result": "retrieved", "sha256": "fdefe8e4854072de0761378c4b1cf3ddf70fe4a7d2d80f40ad7771e3581556ab" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09375", "sha256": "07ef3c83e1ece7b98e40372b12d83e4d9e2e9a813fc695ce5e2fe5d1edd06dd5", "local_path": ".cache/oct12-scan/fulltext/2610.09375.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 03:30:22 UTC (47 KB)", "published": "2026-10-07T03:30:22Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09374", "title": "DUDA-Bench: Benchmarking LLM Agents on Multimodal Data-Driven Urban Diagnosis", "published": "2026-10-07T03:29:27Z", "reviews": [ { "id": "2610.09374", "title": "DUDA-Bench: Benchmarking LLM Agents on Multimodal Data-Driven Urban Diagnosis", "published": "2026-10-07T03:29:27Z", "decision": "reviewed_deferred", "reason": "已读DUDA-Bench method/experiment：86atomic22workflow、12city multimodal十年data、tool/code交互、7model5systems，证据缺口跨阶段传播。下一批评测infra需实际数据分析和验证，atomic正确不等端到端诊断。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 DUDA-Bench", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09374v1", "sha256": "4262cc1235403a671530c24fc2afdabe222c3ec1d1a29238c9ccd65f85d67dd0", "local_html_sha256": "4262cc1235403a671530c24fc2afdabe222c3ec1d1a29238c9ccd65f85d67dd0", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09374v1", "result": "retrieved", "sha256": "4262cc1235403a671530c24fc2afdabe222c3ec1d1a29238c9ccd65f85d67dd0" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09371", "title": "The Confidence Game: Strategic Miscalibration in Human-AI Delegation", "published": "2026-10-07T03:25:44Z", "reviews": [ { "id": "2610.09371", "title": "The Confidence Game: Strategic Miscalibration in Human-AI Delegation", "published": "2026-10-07T03:25:44Z", "decision": "reviewed_deferred", "reason": "已读ConfidenceGame theory/LLMexperiment：repeateddelegation incentives，toyagent明确给true successprob隔离strategicreport，另real-task自估。下一批区分privilegedconfidence诊断与真实uncertainty calibration，不能用toy经济收益当生产效应。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "5 LLM Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 The Confidence Game", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09371v1", "sha256": "2245186060bf120e2d29017ee4c05446a78c783b56facc291baaa9d66c072168", "local_html_sha256": "2245186060bf120e2d29017ee4c05446a78c783b56facc291baaa9d66c072168", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09371v1", "result": "retrieved", "sha256": "2245186060bf120e2d29017ee4c05446a78c783b56facc291baaa9d66c072168" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10608", "title": "From Investigation Failures to Reliable SOC Agents: Understanding and Improving LLM-Based Alert Triage", "published": "2026-10-07T03:20:51Z", "reviews": [ { "id": "2610.10608", "title": "From Investigation Failures to Reliable SOC Agents: Understanding and Improving LLM-Based Alert Triage", "published": "2026-10-07T03:20:51Z", "decision": "reviewed_deferred", "reason": "Google附属机构优先。已读SOC/Aida设计/评测：Strategist提交证据thesis、确定性Executor分辨空与失败、Adversary独立证据、Judge独立终判与有限继续；170开发与1247 AIT-LDS eval隔离。受控alert dataset/F1/FNR/cost结果不是生产SOC A/B。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "V Design of Aida", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 5000 ] }, { "heading": "VI Evaluation of Aida", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 5000 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10608v1", "sha256": "b7f330c4aac0b6f15b4bbad2dcc16ab33dc0f867466c40db3c3a27640362feee", "local_html_sha256": "b7f330c4aac0b6f15b4bbad2dcc16ab33dc0f867466c40db3c3a27640362feee", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10608v1", "result": "retrieved", "sha256": "b7f330c4aac0b6f15b4bbad2dcc16ab33dc0f867466c40db3c3a27640362feee" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09360", "title": "TopoGraphRAG-Bench: Evaluating Multimodal GraphRAG on Layout-Grounded Evidence Reasoning", "published": "2026-10-07T03:15:31Z", "reviews": [ { "id": "2610.09360", "title": "TopoGraphRAG-Bench: Evaluating Multimodal GraphRAG on Layout-Grounded Evidence Reasoning", "published": "2026-10-07T03:15:31Z", "decision": "reviewed_deferred", "reason": "TopoGraphRAG201doc2024QA三evidencetopology counterfactualvalidation；六RAG系统不同familybackends，非纯retrieval单因素比较，需真实layout索引。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "3 Benchmark", "location": ".cache/oct12-scan/fulltext/2610.09360.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "5 Experiments", "section_index": 4, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 227, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09360v1", "sha256": "baba944c64c30558e2ff60c7a50b37342e6e5cafc9a300561df46bc1090bafff", "local_path": ".cache/oct12-scan/fulltext/2610.09360.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09360v1", "result": "retrieved", "sha256": "baba944c64c30558e2ff60c7a50b37342e6e5cafc9a300561df46bc1090bafff" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09360", "sha256": "59e0043898a04392bf2adaf61fa1e464e1b4f32dc41ebdd2d62ec2ba7c66f830", "local_path": ".cache/oct12-scan/fulltext/2610.09360.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 03:15:31 UTC (1,526 KB)", "published": "2026-10-07T03:15:31Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09355", "title": "Multimodal LLMs Can Learn to Read Brain Signals: A Vision--Language Model for Unified Multi-Task EEG Decoding", "published": "2026-10-07T03:13:03Z", "reviews": [ { "id": "2610.09355", "title": "Multimodal LLMs Can Learn to Read Brain Signals: A Vision--Language Model for Unified Multi-Task EEG Decoding", "published": "2026-10-07T03:13:03Z", "decision": "reviewed_deferred", "reason": "BraVista把EEG时频结构图作为VLM接口进行joint multitask后训练，HMC/SEED/EEGMAT/TUAB三seed评测并排除SEED预训练污染；下一批需真实EEG preprocessing/VLM，医疗数据准入与subject隔离。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "2 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Experiments and Results", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09355v1", "sha256": "40671903f841bf1fb498f7517e69a8469952e4a1a9529e6c89d103800981ddd6", "local_html_sha256": "40671903f841bf1fb498f7517e69a8469952e4a1a9529e6c89d103800981ddd6", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09355v1", "result": "retrieved", "sha256": "40671903f841bf1fb498f7517e69a8469952e4a1a9529e6c89d103800981ddd6" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09348", "title": "Relevance Is Not Sufficiency: What Actually Closes the Evidence Gap in Long-Term Memory QA", "published": "2026-10-07T03:09:08Z", "reviews": [ { "id": "2610.09348", "title": "Relevance Is Not Sufficiency: What Actually Closes the Evidence Gap in Long-Term Memory QA", "published": "2026-10-07T03:09:08Z", "decision": "reviewed_deferred", "reason": "已读BFR sufficiency定义/experiment：FCA-MS requirements集合cover后budget unseenrecord acquisition，fixedflatstore same-store控制；GoldHit/TurnHit只coverageproxy，blindedjudgesufficiency另计。下一批被评retriever不得读goldsupport或answer，需真检索与集合补证。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "5 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 WHAT COUNTS AS A SUFFICIENT CANDIDATE MEMORY SET", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09348v1", "sha256": "31c75aa37ef1899b93c1d3fdf0888f4a73d29fe2d55cc7d51f256a4b8ea7a9f7", "local_html_sha256": "31c75aa37ef1899b93c1d3fdf0888f4a73d29fe2d55cc7d51f256a4b8ea7a9f7", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09348v1", "result": "retrieved", "sha256": "31c75aa37ef1899b93c1d3fdf0888f4a73d29fe2d55cc7d51f256a4b8ea7a9f7" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09346", "title": "OnlineQAT: On-Policy Distillation for Ultra-Low-Bit Large Language Models", "published": "2026-10-07T03:05:42Z", "reviews": [ { "id": "2610.09346", "title": "OnlineQAT: On-Policy Distillation for Ultra-Low-Bit Large Language Models", "published": "2026-10-07T03:05:42Z", "decision": "reviewed_deferred", "reason": "OnlineQAT先blockwise W2/W3 QAT再quantized student prefixes OPD恢复，Qwen3-1.7B六benchmark明确group quantization/frozen teacher；下一批需实际低bit forward和两阶段优化，不能浮点假quantization标签。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09346v1", "sha256": "93d2d8dcc7571d9e0017da89ccf5c3d6b0d9f116369c68da21c9c7b82d155093", "local_html_sha256": "93d2d8dcc7571d9e0017da89ccf5c3d6b0d9f116369c68da21c9c7b82d155093", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09346v1", "result": "retrieved", "sha256": "93d2d8dcc7571d9e0017da89ccf5c3d6b0d9f116369c68da21c9c7b82d155093" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09342", "title": "Shared Low-rank Basis Factorization for Data-free Mixture-of-Experts Compression", "published": "2026-10-07T03:01:19Z", "reviews": [ { "id": "2610.09342", "title": "Shared Low-rank Basis Factorization for Data-free Mixture-of-Experts Compression", "published": "2026-10-07T03:01:19Z", "decision": "reviewed_deferred", "reason": "SLBF sharedlowrankexpertbasis与gaugefix保留routing/expertidentity，五16B-122BMoE；data-freeweightfactor真checkpoint重构与heldout任务/GPU待核。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "4 Method", "location": ".cache/oct12-scan/fulltext/2610.09342.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09342.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 228, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09342v1", "sha256": "ea90807f808adf44b35b1ea29cc111917039a901b2e4c0ce98b491128c81fd51", "local_path": ".cache/oct12-scan/fulltext/2610.09342.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09342v1", "result": "retrieved", "sha256": "ea90807f808adf44b35b1ea29cc111917039a901b2e4c0ce98b491128c81fd51" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09342", "sha256": "d61a4f772fe6ed086e0b926da7fd5c9174d0137c85bae5bc7f5c97bcee245eae", "local_path": ".cache/oct12-scan/fulltext/2610.09342.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 03:01:19 UTC (139 KB)", "published": "2026-10-07T03:01:19Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09337", "title": "Ask the Expert: LLM-Guided Reinforcement Learning for Autonomous Cyber Defense", "published": "2026-10-07T02:53:26Z", "reviews": [ { "id": "2610.09337", "title": "Ask the Expert: LLM-Guided Reinforcement Learning for Autonomous Cyber Defense", "published": "2026-10-07T02:53:26Z", "decision": "reviewed_deferred", "reason": "AskExpert LLM对hardcyberdefensestates给受限host建议转换tieredrewardshaping训PPO，deployment不依赖LLM，CAGE1/2benigndefensebenchmark；下一批需可审计限定actions与真实simrollouts，属于LLMguidance非语言权重训练。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "IV Ask the Expert Framework", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "V Experimental Evaluation", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09337v1", "sha256": "60160fae8bf8bb76cdbd0f47a7022977e7b21c311fdcd21104ba21f37f580c65", "local_html_sha256": "60160fae8bf8bb76cdbd0f47a7022977e7b21c311fdcd21104ba21f37f580c65", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09337v1", "result": "retrieved", "sha256": "60160fae8bf8bb76cdbd0f47a7022977e7b21c311fdcd21104ba21f37f580c65" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09328", "title": "Visual Jev Rewards: Reference-Bound Verification for Multi-Subject Image Generation", "published": "2026-10-07T02:34:09Z", "reviews": [ { "id": "2610.09328", "title": "Visual Jev Rewards: Reference-Bound Verification for Multi-Subject Image Generation", "published": "2026-10-07T02:34:09Z", "decision": "reviewed_deferred", "reason": "Visual Jev把reference identities和condition联合Yes标签训练Qwen3.5 verifier后GRPO多主体生成；手工selected897 tasks且单GRPO run/human差异未显著，下一批只能复现参考绑定reward与诊断，不扩大稳定能力结论。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Reference-Bound Visual Jev Rewards", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Data and Experimental Setup", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09328v1", "sha256": "bafd194495e686966ab92efa108b85e7d8987ef44d50ded4de852724d269c946", "local_html_sha256": "bafd194495e686966ab92efa108b85e7d8987ef44d50ded4de852724d269c946", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09328v1", "result": "retrieved", "sha256": "bafd194495e686966ab92efa108b85e7d8987ef44d50ded4de852724d269c946" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.10604", "title": "Beyond Type-checking: Towards Holistic Evaluation of Formal Specification Generation", "published": "2026-10-07T02:11:30Z", "reviews": [ { "id": "2610.10604", "title": "Beyond Type-checking: Towards Holistic Evaluation of Formal Specification Generation", "published": "2026-10-07T02:11:30Z", "decision": "reviewed_deferred", "reason": "已读SpecGen benchmark/results：Lean formalvalidity、referencesimilarity/equivalence、requiredinput/output adequacy分开，350task coverage-aware denominator。100%postconditiontest却0%requiredinput诊断重要，下一批不得用弱precondition刷proofpass或由goldspec生成candidate。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 Benchmark Design", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Experimental Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10604v1", "sha256": "fbee5681c239c92f3626bbd27077f8b1e2c4275ab9bc7ff1c68c4b11a6d6649d", "local_html_sha256": "fbee5681c239c92f3626bbd27077f8b1e2c4275ab9bc7ff1c68c4b11a6d6649d", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10604v1", "result": "retrieved", "sha256": "fbee5681c239c92f3626bbd27077f8b1e2c4275ab9bc7ff1c68c4b11a6d6649d" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09313", "title": "Why VLMs Miss Small Objects, and When Zooming In Is Safe", "published": "2026-10-07T02:10:52Z", "reviews": [ { "id": "2610.09313", "title": "Why VLMs Miss Small Objects, and When Zooming In Is Safe", "published": "2026-10-07T02:10:52Z", "decision": "reviewed_deferred", "reason": "小对象S/L接口几何frontier及nozoomout/overlap分解，177krequests受控/多图像；四ordering失败vendor路径与posthoc必须保留。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "4 Experimental Setup", "location": ".cache/oct12-scan/fulltext/2610.09313.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 Theory", "section_index": 2, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 229, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09313v1", "sha256": "f3af0aca86ca36cbb944043d194155b7ae2f6a86a9612bc29ead32e202ea72a0", "local_path": ".cache/oct12-scan/fulltext/2610.09313.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09313v1", "result": "retrieved", "sha256": "f3af0aca86ca36cbb944043d194155b7ae2f6a86a9612bc29ead32e202ea72a0" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09313", "sha256": "66bebbbf2863170d7ab3465e74b2a4874324fba5e11ad7de2118b23d41848460", "local_path": ".cache/oct12-scan/fulltext/2610.09313.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 02:10:52 UTC (332 KB)", "published": "2026-10-07T02:10:52Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09311", "title": "Denoising Blocks, Not Tokens: Efficient Compressed Continuous Diffusion with Branching Token Realization", "published": "2026-10-07T02:08:22Z", "reviews": [ { "id": "2610.09311", "title": "Denoising Blocks, Not Tokens: Efficient Compressed Continuous Diffusion with Branching Token Realization", "published": "2026-10-07T02:08:22Z", "decision": "reviewed_deferred", "reason": "BLD token→blocklatent、groupedcontinuousdiffusion和parallelARbranches，OpenWebText同GPU成本；长程coherence仍弱，真实生成训练非heuristic分块。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "3 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09311.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "2 Branching Latent Diffusion", "section_index": 1, "characters_read": [ 0, 1401 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 230, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09311v1", "sha256": "7a0d3c99a7d7a1b84c65eb4104c6317330f505aceb888c9bdb76b54857770cee", "local_path": ".cache/oct12-scan/fulltext/2610.09311.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09311v1", "result": "retrieved", "sha256": "7a0d3c99a7d7a1b84c65eb4104c6317330f505aceb888c9bdb76b54857770cee" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09311", "sha256": "70ad61773b00d1642f073e532d28bab72df7352281134c5a6d6c5fc4bb79f591", "local_path": ".cache/oct12-scan/fulltext/2610.09311.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 02:08:22 UTC (310 KB)", "published": "2026-10-07T02:08:22Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09296", "title": "The AI Evaluation Ecosystem", "published": "2026-10-07T01:58:25Z", "reviews": [ { "id": "2610.09296", "title": "The AI Evaluation Ecosystem", "published": "2026-10-07T01:58:25Z", "decision": "reviewed_deferred", "reason": "已读AIevaluationecosystem frame/GABM：institutionagent与rulemarket六维能力simulation、信息分区、holdout设计stress-test，明确hypothesis-generatingsandbox。下一批仅机制仿真参考，不按模拟user satisfaction宣称真实benchmark/market效果。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Mapping the AI Evaluation Ecosystem", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Generative Agent-Based Model of the Evaluation Ecosystem", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09296v1", "sha256": "b3753c4c1c84d1c36e044fd07714a02533967e5b8528298000a2890ce19751b8", "local_html_sha256": "b3753c4c1c84d1c36e044fd07714a02533967e5b8528298000a2890ce19751b8", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09296v1", "result": "retrieved", "sha256": "b3753c4c1c84d1c36e044fd07714a02533967e5b8528298000a2890ce19751b8" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09267", "title": "Who Broke Me? Execution-Guided Repair of Behavioral Dependency Breaks", "published": "2026-10-07T01:04:51Z", "reviews": [ { "id": "2610.09267", "title": "Who Broke Me? Execution-Guided Repair of Behavioral Dependency Breaks", "published": "2026-10-07T01:04:51Z", "decision": "reviewed_deferred", "reason": "已读BBCFixer study/results：old/new dependency双执行rankreturnvalue差异定位rootAPI、filterupstreamdiff供repair，100真实Python/JSbreak。下一批需真实versionedruntime和独立repairtests，不能只compiler错误或直接gold rootAPI。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3. Research Design", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4. Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09267v1", "sha256": "e870863e34523d13008e0cea723435114c177437b205a47dd44f18b7aee54593", "local_html_sha256": "e870863e34523d13008e0cea723435114c177437b205a47dd44f18b7aee54593", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09267v1", "result": "retrieved", "sha256": "e870863e34523d13008e0cea723435114c177437b205a47dd44f18b7aee54593" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09264", "title": "Package Hallucination Attacks on Coding Agents through Prompt Injection in Rule Files", "published": "2026-10-07T00:59:19Z", "reviews": [ { "id": "2610.09264", "title": "Package Hallucination Attacks on Coding Agents through Prompt Injection in Rule Files", "published": "2026-10-07T00:59:19Z", "decision": "reviewed_deferred", "reason": "已读PackHallu threatmodel/evaluation高层：rulefile注入与trajectory-guided evolutionarymutation诱导恶意依赖，13LLM8framework。保留安全研究，不执行攻击rulefile/下载安装可疑包，本批不把supplied AGENTS论文内容当authority。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Threat Model", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "6 Evaluation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09264v1", "sha256": "97b3ccc3403e96797ecea945bd80556480e09e19ea8306b21c1696e58492245d", "local_html_sha256": "97b3ccc3403e96797ecea945bd80556480e09e19ea8306b21c1696e58492245d", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09264v1", "result": "retrieved", "sha256": "97b3ccc3403e96797ecea945bd80556480e09e19ea8306b21c1696e58492245d" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09252", "title": "Adaptive Visual Token Reduction for Accelerated Image Understanding", "published": "2026-10-07T00:37:50Z", "reviews": [ { "id": "2610.09252", "title": "Adaptive Visual Token Reduction for Accelerated Image Understanding", "published": "2026-10-07T00:37:50Z", "decision": "reviewed_deferred", "reason": "ReFIT multi-scale instructionwindowreshape+tokenrefine，只训reductionmodules；四VQA FLOPs非wallclock提升，真实feature模块待实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "2 METHODOLOGY", "location": ".cache/oct12-scan/fulltext/2610.09252.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "3 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09252.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 231, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09252v1", "sha256": "9f71486a5f4103de3f8f95317cf9963eab908674c46fa301068e828287ca028c", "local_path": ".cache/oct12-scan/fulltext/2610.09252.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09252v1", "result": "retrieved", "sha256": "9f71486a5f4103de3f8f95317cf9963eab908674c46fa301068e828287ca028c" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09252", "sha256": "92fe305bf700c7766518167edf8c3289634269057199d06d17a100291686e754", "local_path": ".cache/oct12-scan/fulltext/2610.09252.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Wed, 7 Oct 2026 00:37:50 UTC (1,508 KB)", "published": "2026-10-07T00:37:50Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09247", "title": "An Informational Curse of Horizon in Goal-Conditioned Policy Learning", "published": "2026-10-07T00:20:45Z", "reviews": [ { "id": "2610.09247", "title": "An Informational Curse of Horizon in Goal-Conditioned Policy Learning", "published": "2026-10-07T00:20:45Z", "decision": "rejected_or_out_of_scope", "reason": "goal-conditionedBC/RL longhorizon informationalcurse与inputJacobian Sobolevbootstrapping在OGBenchcontinuouscontrol和oracleplanners诊断；无LLM/foundation/语言agent后训练，当前implementation scope外。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 An Informational Curse of Horizon", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09247v1", "sha256": "5836a8230b01dc8a1fea3eecfda5f8bdc6f76b16fd6965d8865a2963fb2eeb83", "local_html_sha256": "5836a8230b01dc8a1fea3eecfda5f8bdc6f76b16fd6965d8865a2963fb2eeb83", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09247v1", "result": "retrieved", "sha256": "5836a8230b01dc8a1fea3eecfda5f8bdc6f76b16fd6965d8865a2963fb2eeb83" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.09243", "title": "We Query, Therefore We Compute: On Oracle Computation beyond the Machine, with an Application to Agents", "published": "2026-10-07T00:14:09Z", "reviews": [ { "id": "2610.09243", "title": "We Query, Therefore We Compute: On Oracle Computation beyond the Machine, with an Application to Agents", "published": "2026-10-07T00:14:09Z", "decision": "reviewed_deferred", "reason": "已读Oraclemachine/OS theory：two-stack automaton、LLMOracle边界/缓存调度条件与gem5 ArchNights-SE设想，willopensource。理论systems候选保留，尚未读到独立quantitativeagent评测，不宣称runtime集成。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 The Operating System: Two Symmetry Breakings", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "2 The Machine: O-2-PDA", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09243v1", "sha256": "debdbd9e83ae358305f7a60e030bc0a5c40efcccbbca795d679f3e1b6b4e9b58", "local_html_sha256": "debdbd9e83ae358305f7a60e030bc0a5c40efcccbbca795d679f3e1b6b4e9b58", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09243v1", "result": "retrieved", "sha256": "debdbd9e83ae358305f7a60e030bc0a5c40efcccbbca795d679f3e1b6b4e9b58" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09240", "title": "Adversarial Images Hijack Web Agents from Visual Grounding to Browser Execution", "published": "2026-10-07T00:05:53Z", "reviews": [ { "id": "2610.09240", "title": "Adversarial Images Hijack Web Agents from Visual Grounding to Browser Execution", "published": "2026-10-07T00:05:53Z", "decision": "reviewed_deferred", "reason": "white-box局部image扰动按role-slot抽象、page recomposition和action-postprocessing dataflow对齐，离线优化grounding-to-execution而非仅rawresponse。; 四agent配置/六VLM、13公共站点和sandbox多step任务，含defenses和request variation。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.09240v1#S3", "https://arxiv.org/html/2610.09240v1#S4", "https://arxiv.org/html/2610.09240v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09240v1", "sha256": "7f17e1a422f411e1899779c1ddefeac464a3f1e789b5a998deb4bc98e4a415c9" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09227", "title": "Quantize by Drift: Label-Free Mixed-Precision Post-Training Quantization for Text Embedders", "published": "2026-10-06T23:44:32Z", "reviews": [ { "id": "2610.09227", "title": "Quantize by Drift: Label-Free Mixed-Precision Post-Training Quantization for Text Embedders", "published": "2026-10-06T23:44:32Z", "decision": "reviewed_deferred", "reason": "Textembedder label-freemodulequantization输出drift作hardpackedbytebudgetallocation，five development/three sealedheldoutbackbones；mainbudget劣于uniform，下一批必须保negativeresults、真实embedding重算/bitbudget，不能把drift减少当retrieval普遍提升。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "2 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Experimental setup", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09227v1", "sha256": "b677aa9f10d68c9e54377eb1d876b84f5239ce5babbc7410bcf6a0bdc68ea1ee", "local_html_sha256": "b677aa9f10d68c9e54377eb1d876b84f5239ce5babbc7410bcf6a0bdc68ea1ee", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09227v1", "result": "retrieved", "sha256": "b677aa9f10d68c9e54377eb1d876b84f5239ce5babbc7410bcf6a0bdc68ea1ee" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09219", "title": "CM-DPO: Constraint-Margin Direct Preference Optimization for LLM Planning", "published": "2026-10-06T23:30:17Z", "reviews": [ { "id": "2610.09219", "title": "CM-DPO: Constraint-Margin Direct Preference Optimization for LLM Planning", "published": "2026-10-06T23:30:17Z", "decision": "reviewed_deferred", "reason": "补读SynPlan-R及CM-DPO方法：verifiedminimal-edit preference、continuous hardviolation margin与hard满足才启用softobjective，15ktrainingpairs及独立planbench。下一批必须真实severity verifier和DPO梯度，验证/test split与contamination细节继续核，不以discretepreference替代。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "20 Reasoning-Mode Base Models", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 379 ] }, { "heading": "6 Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 The SynPlan-R Synthesis Pipeline", "scope": "supplementary bounded method excerpt", "characters_read_at_least": [ 0, 900 ] }, { "heading": "5 Constraint-Margin Preference Optimization", "scope": "supplementary bounded method excerpt", "characters_read_at_least": [ 0, 900 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09219v1", "sha256": "520b380ecf27bd3eac0505c0858ff450c6c78aabcd7eeb0221a5a0fff6455537", "local_html_sha256": "520b380ecf27bd3eac0505c0858ff450c6c78aabcd7eeb0221a5a0fff6455537", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09219v1", "result": "retrieved", "sha256": "520b380ecf27bd3eac0505c0858ff450c6c78aabcd7eeb0221a5a0fff6455537" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09218", "title": "RLDISCOVER: LLM-driven co-evolution of reinforcement learning algorithms", "published": "2026-10-06T23:26:38Z", "reviews": [ { "id": "2610.09218", "title": "RLDISCOVER: LLM-driven co-evolution of reinforcement learning algorithms", "published": "2026-10-06T23:26:38Z", "decision": "reviewed_deferred", "reason": "已读RLDiscover method/experiments：LLMmutate真RLprogram五component，PCE先target后joint、PPE逐阶段training/repeateval筛，26taskalgorithmpairs。下一批必须实际RL训练和sandboxgeneratedcode，363x近零baseline需绝对return与failedlearning解释，不能静态fitnessfixture。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09218v1", "sha256": "6eed60f186bedc1fa02465319a7cdbefa925302f22a5c5ac09daae3e100e2b1c", "local_html_sha256": "6eed60f186bedc1fa02465319a7cdbefa925302f22a5c5ac09daae3e100e2b1c", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09218v1", "result": "retrieved", "sha256": "6eed60f186bedc1fa02465319a7cdbefa925302f22a5c5ac09daae3e100e2b1c" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09217", "title": "A Deterministic Evidence Layer for Vision-Language Autism Screening from Naturalistic Home Video", "published": "2026-10-06T23:25:18Z", "reviews": [ { "id": "2610.09217", "title": "A Deterministic Evidence Layer for Vision-Language Autism Screening from Naturalistic Home Video", "published": "2026-10-06T23:25:18Z", "decision": "reviewed_deferred", "reason": "frozenVLM eventglossarytable/agecalibration→deterministicmedicalWoE scorer，43homeclips三run；servingvariance不等临床通用，独立data/clinicalvalidation待补。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Method", "location": ".cache/oct12-scan/fulltext/2610.09217.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments and Results", "location": ".cache/oct12-scan/fulltext/2610.09217.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 232, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09217v1", "sha256": "ba2ec3331c53c72c68b2a511a8299db5836237de5d9c1b135aecdbf29e308e8e", "local_path": ".cache/oct12-scan/fulltext/2610.09217.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09217v1", "result": "retrieved", "sha256": "ba2ec3331c53c72c68b2a511a8299db5836237de5d9c1b135aecdbf29e308e8e" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09217", "sha256": "560bd0623fb2a3b4b8d2a61089866ee905399042e044b26f4a703d475b340789", "local_path": ".cache/oct12-scan/fulltext/2610.09217.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Tue, 6 Oct 2026 23:25:18 UTC (5,852 KB)", "published": "2026-10-06T23:25:18Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09209", "title": "Multi-Objective Aligned Small Language Model Framework for SUD Patient Dialogue Generation", "published": "2026-10-06T23:10:21Z", "reviews": [ { "id": "2610.09209", "title": "Multi-Objective Aligned Small Language Model Framework for SUD Patient Dialogue Generation", "published": "2026-10-06T23:10:21Z", "decision": "reviewed_deferred", "reason": "SUDdialogue用CCD认知component检测→CADG KD/SFT/preferences/attentionreward，数据来自publicReddit personalnarratives再构patienthistories；下一批需医疗伦理/隐私审查与不臆造临床真实度，不能仅LLMhitmetric当therapy有效。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "III Methodology", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "IV Evaluation", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09209v1", "sha256": "347e6cc18f9c3f316a6cd31ad668e0cb635de16ecf3be48222030e91506ded4d", "local_html_sha256": "347e6cc18f9c3f316a6cd31ad668e0cb635de16ecf3be48222030e91506ded4d", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09209v1", "result": "retrieved", "sha256": "347e6cc18f9c3f316a6cd31ad668e0cb635de16ecf3be48222030e91506ded4d" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09205", "title": "Do Vision Models Learn Physical Constraints or Rendering Shortcuts? A Counterfactual Benchmark for Grounded Physical Consistency", "published": "2026-10-06T23:06:44Z", "reviews": [ { "id": "2610.09205", "title": "Do Vision Models Learn Physical Constraints or Rendering Shortcuts? A Counterfactual Benchmark for Grounded Physical Consistency", "published": "2026-10-06T23:06:44Z", "decision": "reviewed_deferred", "reason": "Mitsuba500scenefamily物理counterfactualbenchmark与LoRA测量baseline；interventionshift明显退化，新architecture非贡献，真renderer和familydisjoint待核。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "5 Benchmark Models and Adaptation", "location": ".cache/oct12-scan/fulltext/2610.09205.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Counterfactual Physical Consistency Benchmark", "location": ".cache/oct12-scan/fulltext/2610.09205.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 233, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09205v1", "sha256": "359571845c66760cd013d68e16f4e80f668dc7d03ae3f2ce82cf3f7e00fa3829", "local_path": ".cache/oct12-scan/fulltext/2610.09205.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09205v1", "result": "retrieved", "sha256": "359571845c66760cd013d68e16f4e80f668dc7d03ae3f2ce82cf3f7e00fa3829" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09205", "sha256": "a9bac037e735ed8a570216f586db24c3c71a4a61bd8b2765191e6c32b6564c09", "local_path": ".cache/oct12-scan/fulltext/2610.09205.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Tue, 6 Oct 2026 23:06:44 UTC (2,210 KB)", "published": "2026-10-06T23:06:44Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09202", "title": "Few Bits, One Law: Toward W2A4KV2", "published": "2026-10-06T22:55:36Z", "reviews": [ { "id": "2610.09202", "title": "Few Bits, One Law: Toward W2A4KV2", "published": "2026-10-06T22:55:36Z", "decision": "reviewed_deferred", "reason": "固定HDP rotation+L2 energy canonicalization，离线Gaussian scalar/vector frozen codebooks，联合W/A/K/V真实QAT；normalization-aware STE不是粗略identity整块gradient。; §5跨LLaMA1/3/8B、Qwen1.7B，WT2/eight zero-shot tasks、8K continuation与generation；零QAT在W2A4的PPL极差，需真实训练。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract/author information/main defining methods/main experiments and tables; production-evidence paragraph spot check; appendices not exhaustively read", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.09202v1#S3", "https://arxiv.org/html/2610.09202v1#S4", "https://arxiv.org/html/2610.09202v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09202v1", "sha256": "4c89751e111611a7e0ec5e9fdbcedd29d24dce7a1855fbdb70a29141324fc446" }, "new_implementation_inferred_from_review": false }, { "id": "2610.09202", "title": "Few Bits, One Law: Toward W2A4KV2", "published": "2026-10-06T22:55:36Z", "decision": "reviewed_deferred", "reason": "Meta CanonQ fixedrotation/L2canonicalization+frozenGaussiancodebook联合W/A/K/VQAT，singletrainseed；GSM8K用于LR选择不能当isolatedtest。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 CanonQ : Unified W/A/KV Quantization", "location": ".cache/oct12-scan/fulltext/2610.09202.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09202.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 234, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09202v1", "sha256": "4c89751e111611a7e0ec5e9fdbcedd29d24dce7a1855fbdb70a29141324fc446", "local_path": ".cache/oct12-scan/fulltext/2610.09202.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09202v1", "result": "retrieved", "sha256": "4c89751e111611a7e0ec5e9fdbcedd29d24dce7a1855fbdb70a29141324fc446" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09202", "sha256": "ab15a1c9deed4101d0b394104f7adc119df3cfd0d8a34ee1c058b78d04610388", "local_path": ".cache/oct12-scan/fulltext/2610.09202.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Tue, 6 Oct 2026 22:55:36 UTC (564 KB)", "published": "2026-10-06T22:55:36Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09193", "title": "Bookkeeping, Composition, or Unreachable Gold? Reading MemoryAgentBench's Conflict-Resolution Scores Against a Frozen Last-Write Resolver", "published": "2026-10-06T22:44:32Z", "reviews": [ { "id": "2610.09193", "title": "Bookkeeping, Composition, or Unreachable Gold? Reading MemoryAgentBench's Conflict-Resolution Scores Against a Frozen Last-Write Resolver", "published": "2026-10-06T22:44:32Z", "decision": "reviewed_deferred", "reason": "MemoryAgentBench frozenlastwriteresolver在1devlist后测三list并67unreachablegold审计；symboliccontrol非agent能力，releasegold只在external评估。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "3 Results", "location": ".cache/oct12-scan/fulltext/2610.09193.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "2 Setup", "section_index": 1, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 235, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09193v1", "sha256": "0e341a87909b86dbbe3dcedb7d2950f7aa78aef0a8036d3b7d4fb265f8d67dd2", "local_path": ".cache/oct12-scan/fulltext/2610.09193.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09193v1", "result": "retrieved", "sha256": "0e341a87909b86dbbe3dcedb7d2950f7aa78aef0a8036d3b7d4fb265f8d67dd2" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09193", "sha256": "f7a8bf01eae4059aa88cbf355d0d7f1decac81510ded6316b0e7e8961c2a33d9", "local_path": ".cache/oct12-scan/fulltext/2610.09193.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Tue, 6 Oct 2026 22:44:32 UTC (53 KB)", "published": "2026-10-06T22:44:32Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09186", "title": "The Dichotomy Between Pattern Recognition and Step-by-Step Reasoning", "published": "2026-10-06T22:36:42Z", "reviews": [ { "id": "2610.09186", "title": "The Dichotomy Between Pattern Recognition and Step-by-Step Reasoning", "published": "2026-10-06T22:36:42Z", "decision": "reviewed_deferred", "reason": "DeBruijn reasoningstate频率理论/训练Qwen1.5与restrictedattentionQwen14/32B；accuracy与扰动fragilitytradeoff，真实stateemit训练待下一批。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "additional targeted method/evaluation", "heading": "3 the degree of de bruijn structure in the training data determines how much an LLM uses pattern recognition or step-by-step reasoning", "section_index": 2, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "4 real-world reasoning traces exhibit de bruijn structure.", "section_index": 3, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 236, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09186v1", "sha256": "c32e1cf4bc50d14bf18678476a58b4cce6bcb3deebfe6db0bf9e12ae15e87e80", "local_path": ".cache/oct12-scan/fulltext/2610.09186.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09186v1", "result": "retrieved", "sha256": "c32e1cf4bc50d14bf18678476a58b4cce6bcb3deebfe6db0bf9e12ae15e87e80" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09186", "sha256": "297b5403ceb8887a62fbf234b05c3e2b81dce319207edb719b7d8d822acd8b34", "local_path": ".cache/oct12-scan/fulltext/2610.09186.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Tue, 6 Oct 2026 22:36:42 UTC (527 KB)", "published": "2026-10-06T22:36:42Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09179", "title": "LayerRoPE: Dynamic Depth-wise Magnitude & Angular Superposition", "published": "2026-10-06T22:30:58Z", "reviews": [ { "id": "2610.09179", "title": "LayerRoPE: Dynamic Depth-wise Magnitude & Angular Superposition", "published": "2026-10-06T22:30:58Z", "decision": "reviewed_deferred", "reason": "LayerRoPE共享gain向量加depth rotation/scalars替换layerwise gamma，C4同顺序scaling/depth/LR/looped模型控制；主要foundation预训练architecture而非OPD，下一批需真实前向与scaled训练。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 LayerRoPE", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments and Results", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09179v1", "sha256": "7a67b989be87aa337de905348072114bc41c0ff742d603007e19e3d042ad506c", "local_html_sha256": "7a67b989be87aa337de905348072114bc41c0ff742d603007e19e3d042ad506c", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09179v1", "result": "retrieved", "sha256": "7a67b989be87aa337de905348072114bc41c0ff742d603007e19e3d042ad506c" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09163", "title": "ToolRACER: A Robust Agentic Conversation Emulation Resource for Agent Training and Evaluation", "published": "2026-10-06T22:06:17Z", "reviews": [ { "id": "2610.09163", "title": "ToolRACER: A Robust Agentic Conversation Emulation Resource for Agent Training and Evaluation", "published": "2026-10-06T22:06:17Z", "decision": "reviewed_deferred", "reason": "已读ToolRACER methodology/training：user/assistant/tool emulation、四judgevalidation及repairresimulation、3.7ktrain vs5kAPIGenMT ablation。下一批需真实smallmodel SFT及分离in-domain混合效应，emulatedtool结果非productiontool成功。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 ToolRACER: Methodology for Robust Synthetic Multi-Turn data Generation", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiment", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09163v1", "sha256": "48934f598e2077a8833cb4e5ceb06af773ef10921f61cc9010c415d3f393c053", "local_html_sha256": "48934f598e2077a8833cb4e5ceb06af773ef10921f61cc9010c415d3f393c053", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09163v1", "result": "retrieved", "sha256": "48934f598e2077a8833cb4e5ceb06af773ef10921f61cc9010c415d3f393c053" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09159", "title": "SpecGuard: Proving a Task Is Broken Before the Agent Cheats", "published": "2026-10-06T22:00:49Z", "reviews": [ { "id": "2610.09159", "title": "SpecGuard: Proving a Task Is Broken Before the Agent Cheats", "published": "2026-10-06T22:00:49Z", "decision": "reviewed_deferred", "reason": "Google DeepMind优先。已读SpecGuard方法、实验和限制：SpecAgent独立测试生成Lean specification，TestAgent分离，semantic connector与kernel证书只有在忠实autoformalization下保证。349 mutated SWE任务和手工60规范审计，不是线上A/B；下一批须防规范弱化/伪冲突。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 SpecGuard", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 3898 ] }, { "heading": "5 Experiments", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 12169 ] }, { "heading": "6 Discussion and Limitations", "scope": "bounded cached section text; source sections can be truncated", "characters_read_at_least": [ 0, 2666 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09159v1", "sha256": "4ae091de142f50c018e524f4455364be0f6b147d95cce4729097987597f91ee6", "local_html_sha256": "4ae091de142f50c018e524f4455364be0f6b147d95cce4729097987597f91ee6", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09159v1", "result": "retrieved", "sha256": "4ae091de142f50c018e524f4455364be0f6b147d95cce4729097987597f91ee6" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09152", "title": "sk-bench: A Native-First Benchmark for Evaluating Large Language Models in Slovak", "published": "2026-10-06T21:51:22Z", "reviews": [ { "id": "2610.09152", "title": "sk-bench: A Native-First Benchmark for Evaluating Large Language Models in Slovak", "published": "2026-10-06T21:51:22Z", "decision": "reviewed_deferred", "reason": "skbench native-first30dataset/33variants统一genharness55models；native/translated/LLMQA provenance区分，Slovakcontinualtrain损失+instructionrepair需真checkpoint。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "4 Evaluation Harness", "location": ".cache/oct12-scan/fulltext/2610.09152.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 The sk-bench Task Suite", "section_index": 2, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 237, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09152v1", "sha256": "d07285b6712e7eb225bd62aacdbb2eed5025e2aea41cf1aec022aecabefa3112", "local_path": ".cache/oct12-scan/fulltext/2610.09152.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09152v1", "result": "retrieved", "sha256": "d07285b6712e7eb225bd62aacdbb2eed5025e2aea41cf1aec022aecabefa3112" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09152", "sha256": "f362dc629c8fbeacd49afbd11289f725dafc11cd67632060d9bef1ab54359f40", "local_path": ".cache/oct12-scan/fulltext/2610.09152.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Tue, 6 Oct 2026 21:51:22 UTC (236 KB)", "published": "2026-10-06T21:51:22Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09142", "title": "Finding Blind Spots in AppWorld and WorkArena Task Verifiers", "published": "2026-10-06T21:34:00Z", "reviews": [ { "id": "2610.09142", "title": "Finding Blind Spots in AppWorld and WorkArena Task Verifiers", "published": "2026-10-06T21:34:00Z", "decision": "reviewed_deferred", "reason": "已读AppWorld/WorkArena audit threat/protocol：不改shippedchecker、source-informed mutation与independentreadback、prospectiveincrement冻结，23selectedcase非populationrate、offdiagonal拒绝cause缺证分开。下一批重要CPUevalinfra，无modelcalls，不能广告agent能力算法。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 Threat Model and Fault-Construction Grammar", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 The Audit", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09142v1", "sha256": "45120bccc21f0bc6a321677907c06d1bce32880cbf942db7aee605448d4665f7", "local_html_sha256": "45120bccc21f0bc6a321677907c06d1bce32880cbf942db7aee605448d4665f7", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09142v1", "result": "retrieved", "sha256": "45120bccc21f0bc6a321677907c06d1bce32880cbf942db7aee605448d4665f7" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09141", "title": "A Cognitive-Aware QML-CRL Framework for Detecting Affinity and Romance-Investment Fraud", "published": "2026-10-06T21:33:26Z", "reviews": [ { "id": "2610.09141", "title": "A Cognitive-Aware QML-CRL Framework for Detecting Affinity and Romance-Investment Fraud", "published": "2026-10-06T21:33:26Z", "decision": "rejected_or_out_of_scope", "reason": "已读quantumencoder/jointRL：PQC cognitivebiasqubits与Wait/Flag optimalstopping传统quantum-classical fraud模型、syntheticconversation，不是LLM agent中心算法，当前范围外。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 Quantum-Cognitive PQC Encoder", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 Joint Quantum-Classical Learning", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09141v1", "sha256": "f1f48397c8e89e3b6c80a6c5f6127081623fe54ad6b20dee91fa5103c9772925", "local_html_sha256": "f1f48397c8e89e3b6c80a6c5f6127081623fe54ad6b20dee91fa5103c9772925", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09141v1", "result": "retrieved", "sha256": "f1f48397c8e89e3b6c80a6c5f6127081623fe54ad6b20dee91fa5103c9772925" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.09139", "title": "Breaking the Space Barrier and its Application to Language Model Inference", "published": "2026-10-06T21:29:54Z", "reviews": [ { "id": "2610.09139", "title": "Breaking the Space Barrier and its Application to Language Model Inference", "published": "2026-10-06T21:29:54Z", "decision": "reviewed_deferred", "reason": "已读structuredinference theory/experiment：fewpathautomata约束token、强制文字免model、onGPU重算mask/recursive stack、AppleM2实际MLXmatcheddecode。下一批foundation/serving候选需要真实grammarvalidity与resource，理论复杂度不直接证明模型能力。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "5 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 The theory in plain words", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09139v1", "sha256": "026328fdaabe303c22b5cd49fc09a7db80c919e94e680dcb6282edc89ecb1455", "local_html_sha256": "026328fdaabe303c22b5cd49fc09a7db80c919e94e680dcb6282edc89ecb1455", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09139v1", "result": "retrieved", "sha256": "026328fdaabe303c22b5cd49fc09a7db80c919e94e680dcb6282edc89ecb1455" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09127", "title": "CADFather: Autonomous CAD Reconstruction through Coordinated Tool Use", "published": "2026-10-06T21:20:59Z", "reviews": [ { "id": "2610.09127", "title": "CADFather: Autonomous CAD Reconstruction through Coordinated Tool Use", "published": "2026-10-06T21:20:59Z", "decision": "reviewed_deferred", "reason": "已读CADFather method/experiment：VLM看target/intermediate renders，协调pretrainedCADgeneration/numericalrefinement，真CadQueryDSLexecute feedback与bestvalidarchive，无新training。下一批须真实3D几何/多tool策略，注意testsubsetablation不能参与parameter选择。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09127v1", "sha256": "fb2f3d65fa62b29b627689158900bea384b0e94aae9d8e8ea960f5d41ae8cc14", "local_html_sha256": "fb2f3d65fa62b29b627689158900bea384b0e94aae9d8e8ea960f5d41ae8cc14", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09127v1", "result": "retrieved", "sha256": "fb2f3d65fa62b29b627689158900bea384b0e94aae9d8e8ea960f5d41ae8cc14" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09120", "title": "The Deceptive Bandit Problem: Exploratory Coupling and the Fragility of Multi-Agent Learning", "published": "2026-10-06T21:12:21Z", "reviews": [ { "id": "2610.09120", "title": "The Deceptive Bandit Problem: Exploratory Coupling and the Fragility of Multi-Agent Learning", "published": "2026-10-06T21:12:21Z", "decision": "rejected_or_out_of_scope", "reason": "已读deceptivebandit理论/prelim：stronglymonotone两playerbandit、探索相关泄漏导致DNE及收敛证明，传统gamelearning非LLMagent，本批范围外，不开展实际数据泄漏攻击。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "V Main Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "II Preliminaries", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09120v1", "sha256": "234fe816694b214cf3d08f083d89b5a372fb3f2d0b56f380a431f68085898b80", "local_html_sha256": "234fe816694b214cf3d08f083d89b5a372fb3f2d0b56f380a431f68085898b80", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09120v1", "result": "retrieved", "sha256": "234fe816694b214cf3d08f083d89b5a372fb3f2d0b56f380a431f68085898b80" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.09115", "title": "From Uncertainty to Action: Learning to Steer LLM Agents", "published": "2026-10-06T21:05:50Z", "reviews": [ { "id": "2610.09115", "title": "From Uncertainty to Action: Learning to Steer LLM Agents", "published": "2026-10-06T21:05:50Z", "decision": "reviewed_deferred", "reason": "已读VoS stepcounterfactual/uncertaintyanalysis：82k真实continuation SOT学何处/何机制steer，harmbudget限制干扰success。下一批需真实value训练与按task隔离SOT、因果续跑而非把realizedoutcome直接作为forecast。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "2 Stepwise Counterfactual Evaluation of Steering", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "3 What Uncertainty Tells a Steering Policy", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09115v1", "sha256": "ad7712a8a364ece54b5911cce845a4a9be7a21cd0f9820cba3d31ce93567d365", "local_html_sha256": "ad7712a8a364ece54b5911cce845a4a9be7a21cd0f9820cba3d31ce93567d365", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09115v1", "result": "retrieved", "sha256": "ad7712a8a364ece54b5911cce845a4a9be7a21cd0f9820cba3d31ce93567d365" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09112", "title": "GeoNatureAgent (GNA): A Framework and Benchmark for Pre-Production Evaluation of Tool-Using Agents on Geospatial and Environmental Tasks", "published": "2026-10-06T21:03:04Z", "reviews": [ { "id": "2610.09112", "title": "GeoNatureAgent (GNA): A Framework and Benchmark for Pre-Production Evaluation of Tool-Using Agents on Geospatial and Environmental Tasks", "published": "2026-10-06T21:03:04Z", "decision": "reviewed_deferred", "reason": "已读GNA framework/benchmark：固定16MCPtool接同selfhostAPI、103task strictallchecksvs percheckpartial、3seed capability/cost分开。下一批preproductioninfra需actualAPI/geometry计算和deterministicscoring，不能候选模型不同tool栈或testgold参数。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3. The GNA Framework", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4. The Benchmark Instance", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09112v1", "sha256": "7117dff1a98c9a1f75c3799466c652a83c8f8903ecdadba04574d3a88ffa6a8e", "local_html_sha256": "7117dff1a98c9a1f75c3799466c652a83c8f8903ecdadba04574d3a88ffa6a8e", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09112v1", "result": "retrieved", "sha256": "7117dff1a98c9a1f75c3799466c652a83c8f8903ecdadba04574d3a88ffa6a8e" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09096", "title": "Are We Really Benchmarking Forecasting Models? The Impact of Preprocessing on Time Series Performance", "published": "2026-10-06T20:50:28Z", "reviews": [ { "id": "2610.09096", "title": "Are We Really Benchmarking Forecasting Models? The Impact of Preprocessing on Time Series Performance", "published": "2026-10-06T20:50:28Z", "decision": "rejected_or_out_of_scope", "reason": "M4专用forecasting预处理metadataset，不含本批LLM/foundation核心算法；不能把per-seriesoraclepipeline优化视为test部署收益。", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "read_scope": "abstract_scope_exclusion; original HTML transport is not counted as reading", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 238, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09096v1", "sha256": "ca937138dd9a98ac2816ad58a3c42ef773d18b5ede4292faf648ab43b286747a", "local_path": ".cache/oct12-scan/fulltext/2610.09096.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09096v1", "result": "retrieved", "sha256": "ca937138dd9a98ac2816ad58a3c42ef773d18b5ede4292faf648ab43b286747a" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09096", "sha256": "31755ef8a4f4124f5ef0649b2a55468aaafb2659a2ac084a37ec96ead7e71c5d", "local_path": ".cache/oct12-scan/fulltext/2610.09096.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Tue, 6 Oct 2026 20:50:28 UTC (1,432 KB)", "published": "2026-10-06T20:50:28Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.09085", "title": "Anaximander: Interactively Running Geospatial Deep Learning Models on Any Compute Backend", "published": "2026-10-06T20:31:38Z", "reviews": [ { "id": "2610.09085", "title": "Anaximander: Interactively Running Geospatial Deep Learning Models on Any Compute Backend", "published": "2026-10-06T20:31:38Z", "decision": "rejected_or_out_of_scope", "reason": "Anaximander QGIS/gRPC多backend集成/农业segmentation演示，无新LLM/foundation算法或受控capability研究，本批范围外。", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "read_scope": "abstract_scope_exclusion; original HTML transport is not counted as reading", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 239, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09085v1", "sha256": "fa39a81b6e6c63173b8708a132296c9630244738ceea1824d052384c705880f7", "local_path": ".cache/oct12-scan/fulltext/2610.09085.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09085v1", "result": "retrieved", "sha256": "fa39a81b6e6c63173b8708a132296c9630244738ceea1824d052384c705880f7" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09085", "sha256": "0db682b44cf0ff3edd0b7e28da643b83c1e62b3f9ef3452e59343e4f02f0575c", "local_path": ".cache/oct12-scan/fulltext/2610.09085.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Tue, 6 Oct 2026 20:31:38 UTC (7,744 KB)", "published": "2026-10-06T20:31:38Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "rejected_or_out_of_scope" }, { "id": "2610.09079", "title": "Large-scale Repository Engineering via Agent-Native Reusable Code Primitives", "published": "2026-10-06T20:20:50Z", "reviews": [ { "id": "2610.09079", "title": "Large-scale Repository Engineering via Agent-Native Reusable Code Primitives", "published": "2026-10-06T20:20:50Z", "decision": "reviewed_deferred", "reason": "CodeFace primitive携带interface/dependency/tests/provenance且residentLLM适配；LEGO做cross-componentconstraint resolution和真实executedtest诊断定点再激活，budget5/module。; 522 packages、13backbones、真实native suites空packagefloor到original-sourceceiling；有remine/disjoint控制与test-special-casing抽样审查。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.09079v1#S3", "https://arxiv.org/html/2610.09079v1#S4", "https://arxiv.org/html/2610.09079v1#S5" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09079v1", "sha256": "559a07bff49ae267844b7d1c56e4de5c78787c8e7e64dd29c6340b0c70b60d13" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09074", "title": "TAP: Efficient Long-Horizon Agent Pruning via Trajectory-Anchored Recovery", "published": "2026-10-06T20:18:17Z", "reviews": [ { "id": "2610.09074", "title": "TAP: Efficient Long-Horizon Agent Pruning via Trajectory-Anchored Recovery", "published": "2026-10-06T20:18:17Z", "decision": "reviewed_deferred", "reason": "TAP iterative FFN structural pruning结合teacher-trajectory anchored student on-policy responses，恢复loss梯度重评channel；ALFWorld/WebShop pruning/latency评测，需真实剪除并恢复train/infer，下一批。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09074v1", "sha256": "b639a1033aef4c92200424599dcb57965ed8d85701a2170cb612540caacad1a5", "local_html_sha256": "b639a1033aef4c92200424599dcb57965ed8d85701a2170cb612540caacad1a5", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09074v1", "result": "retrieved", "sha256": "b639a1033aef4c92200424599dcb57965ed8d85701a2170cb612540caacad1a5" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.09070", "title": "Covariate-dependent Joint Modeling of Multivariate Ordinal Preferences and Its Connections with Comparison Models", "published": "2026-10-06T20:15:43Z", "reviews": [ { "id": "2610.09070", "title": "Covariate-dependent Joint Modeling of Multivariate Ordinal Preferences and Its Connections with Comparison Models", "published": "2026-10-06T20:15:43Z", "decision": "reviewed_deferred", "reason": "CR–MRF joint ordinal preference统计方法有MovieLens与合成MLE/MPLE对照，但§7明确LLM reward model只是future work。没有真正LLM/工业生产训练链路；保留统计参考，不伪装后训练算法已实现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "bounded_methods_and_evaluation", "review_artifact": "research-audits/oct12-review-closure/recommendation-reviews.json", "evidence_locations": [ "§3.4 likelihood estimation", "§5 Numerical experiments", "§7 Conclusions", "Supp S.6" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09070v1", "sha256": "34abdcd7f076570064aaf6302f7a705485ff3cf851639565eae908eddd61bff0" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "recommendation" ], "status": "reviewed_deferred" }, { "id": "2610.09063", "title": "Multi-Label Topic Assignment via LLM Distillation: A Comparative Analysis of Generative vs. Discriminative Student Models", "published": "2026-10-06T20:12:11Z", "reviews": [ { "id": "2610.09063", "title": "Multi-Label Topic Assignment via LLM Distillation: A Comparative Analysis of Generative vs. Discriminative Student Models", "published": "2026-10-06T20:12:11Z", "decision": "reviewed_deferred", "reason": "teacherLLM标内部marketplace topics后比较SLM/DeBERTa/ModernBERT；原文§5有250QPS/9H100及20%rampAB获批review+60%，但私有数据阻止公开复现。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "2 Methodology: Scalable Multi-Label Classification", "location": ".cache/oct12-scan/fulltext/2610.09063.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "3 Experimental Setup and Use Cases", "location": ".cache/oct12-scan/fulltext/2610.09063.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "5 Impact and Production Deployment", "section_index": 4, "characters_read": [ 0, 1330 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 240, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09063v1", "sha256": "a11a208e830d46f37dfc32a8f3b920c912a1b88eda6c368ef6851d18042df3ac", "local_path": ".cache/oct12-scan/fulltext/2610.09063.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09063v1", "result": "retrieved", "sha256": "a11a208e830d46f37dfc32a8f3b920c912a1b88eda6c368ef6851d18042df3ac" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09063", "sha256": "17af441d38fa83d86d3bfb37776c61bc3c594f25c8be8481179668ffa07b399e", "local_path": ".cache/oct12-scan/fulltext/2610.09063.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Tue, 6 Oct 2026 20:12:11 UTC (50 KB)", "published": "2026-10-06T20:12:11Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09053", "title": "Justice After Identity: Large Language Models and the View from Everywhere", "published": "2026-10-06T20:01:59Z", "reviews": [ { "id": "2610.09053", "title": "Justice After Identity: Large Language Models and the View from Everywhere", "published": "2026-10-06T20:01:59Z", "decision": "reviewed_deferred", "reason": "已读JusticeAfterIdentity完整摘要，官方正文未取得：哲学originalposition与identityconstraint的LLM/human moraljudgment数据，主要讨论性研究。未核验methods/metric，保留unresolved而不假称fulltext排除。", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "read_scope": "abstract_only_primary_fulltext_unavailable", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [], "dateproof": null, "source_proof": { "url": "https://arxiv.org/abs/2610.09053", "sha256": null, "local_html_sha256": null, "local_html_sha256_matches": false, "attempts": [ { "url": "https://arxiv.org/html/2610.09053v1", "result": "failed", "error": "404 Client Error: Not Found for url: https://arxiv.org/html/2610.09053v1" }, { "url": "https://arxiv.org/html/2610.09053v1", "result": "failed", "error": "404 Client Error: Not Found for url: https://arxiv.org/html/2610.09053v1" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09051", "title": "A Self-Pruning Transformer: Extreme KV-Cache Compression with Universal Attention", "published": "2026-10-06T20:00:04Z", "reviews": [ { "id": "2610.09051", "title": "A Self-Pruning Transformer: Extreme KV-Cache Compression with Universal Attention", "published": "2026-10-06T20:00:04Z", "decision": "reviewed_deferred", "reason": "UniversalAttention learnedcontentdecay保留RoPE/softmax并adaptiveKVprune，1B50BDolma/RULER；真实attention训练和prune kernel待实施。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "4 Universal Attention", "location": ".cache/oct12-scan/fulltext/2610.09051.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5 Evaluation", "location": ".cache/oct12-scan/fulltext/2610.09051.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 241, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09051v1", "sha256": "e52ebc15f9e103c03e4d0c62473e8203cb4acf2a339168c26835e6f51d41dd74", "local_path": ".cache/oct12-scan/fulltext/2610.09051.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09051v1", "result": "retrieved", "sha256": "e52ebc15f9e103c03e4d0c62473e8203cb4acf2a339168c26835e6f51d41dd74" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09051", "sha256": "4b90fe59417b32646101be33f15a7d856567c33a94507b33854dd3467693e5fd", "local_path": ".cache/oct12-scan/fulltext/2610.09051.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Tue, 6 Oct 2026 20:00:04 UTC (1,592 KB)", "published": "2026-10-06T20:00:04Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09027", "title": "Visual Memory Attacks Can Persist Through The KV Cache", "published": "2026-10-06T19:25:08Z", "reviews": [ { "id": "2610.09027", "title": "Visual Memory Attacks Can Persist Through The KV Cache", "published": "2026-10-06T19:25:08Z", "decision": "reviewed_deferred", "reason": "P-VMI whitebox图像攻击移除后仍经后续KV影响，cache-swap/summarycompaction控制；真KV交互/梯度攻击待实现，onlysoftprompt是diagnostic。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09027.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 Persistent attacks", "section_index": 2, "characters_read": [ 0, 1400 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 242, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09027v1", "sha256": "464b886c7604a003e18de48066ba90fcb89d7d93444a4ce9d54057edb2b9ff03", "local_path": ".cache/oct12-scan/fulltext/2610.09027.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09027v1", "result": "retrieved", "sha256": "464b886c7604a003e18de48066ba90fcb89d7d93444a4ce9d54057edb2b9ff03" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09027", "sha256": "55ef3fc59de260a6b01b32bccc82f645c25f6b33f0bacd5b457ee479db34ae4a", "local_path": ".cache/oct12-scan/fulltext/2610.09027.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Tue, 6 Oct 2026 19:25:08 UTC (1,263 KB)", "published": "2026-10-06T19:25:08Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09021", "title": "Not Every Call Needs a Frontier Model: Per-Call-Site Evaluation of Small Language Models in a Deployed Agentic Home-Automation System", "published": "2026-10-06T19:18:59Z", "reviews": [ { "id": "2610.09021", "title": "Not Every Call Needs a Frontier Model: Per-Call-Site Evaluation of Small Language Models in a Deployed Agentic Home-Automation System", "published": "2026-10-06T19:18:59Z", "decision": "reviewed_deferred", "reason": "已读Wactorz callsites/setup/results：真实deployedprompt与两HAregistry、280case2520calls，intent/actuation/planner/codegen不同能力，错误设备actuation/refusal单记。liveuser43case是小部署比较非随机生产AB已证，下一批仅批准localmock/安全device模拟，不控制用户家居。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Call sites, benchmark and setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09021v1", "sha256": "ac02df6985dfcc06f0036c7fb0386c5f4fefc0cb087f5b8b0f6d0da29a6dc6d3", "local_html_sha256": "ac02df6985dfcc06f0036c7fb0386c5f4fefc0cb087f5b8b0f6d0da29a6dc6d3", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09021v1", "result": "retrieved", "sha256": "ac02df6985dfcc06f0036c7fb0386c5f4fefc0cb087f5b8b0f6d0da29a6dc6d3" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09015", "title": "Personalize at Test Time: Learning User Preferences for Image Generation", "published": "2026-10-06T19:10:08Z", "reviews": [ { "id": "2610.09015", "title": "Personalize at Test Time: Learning User Preferences for Image Generation", "published": "2026-10-06T19:10:08Z", "decision": "reviewed_deferred", "reason": "Meta PAT attributeAE50维evaluator+BradleyTerry per-user/mixedeffectweights，631users×50heldout/三split；真实diffusionrewardguidance与toy/realuser端点分开。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Methodology", "location": ".cache/oct12-scan/fulltext/2610.09015.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.09015.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "3 Methodology", "section_index": 2, "characters_read": [ 0, 1801 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "4 Experiments", "section_index": 3, "characters_read": [ 0, 1800 ], "scope": "bounded excerpt; original source SHA verified" }, { "role": "additional targeted method/evaluation", "heading": "4 Experiments", "location": "text after 4.2 Real User: Personalized Weight Modeling", "characters_read": [ 0, 2400 ], "scope": "bounded excerpt; raw source available, cached section extraction may be capped" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 243, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09015v1", "sha256": "8b5b609aa9c7cd0db5bbbad494004d1b624530adc1ce5bbffd0449ae9fc01be7", "local_path": ".cache/oct12-scan/fulltext/2610.09015.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09015v1", "result": "retrieved", "sha256": "8b5b609aa9c7cd0db5bbbad494004d1b624530adc1ce5bbffd0449ae9fc01be7" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09015", "sha256": "fe59f750344f67a07e363b641c7a5f77570b4b93c0eb4bb15b5c391e52149337", "local_path": ".cache/oct12-scan/fulltext/2610.09015.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Tue, 6 Oct 2026 19:10:08 UTC (25,987 KB)", "published": "2026-10-06T19:10:08Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09008", "title": "Whose Memory Is It? Scope-Aware Commit Rules for Long-Term LLM Memory", "published": "2026-10-06T19:05:31Z", "reviews": [ { "id": "2610.09008", "title": "Whose Memory Is It? Scope-Aware Commit Rules for Long-Term LLM Memory", "published": "2026-10-06T19:05:31Z", "decision": "reviewed_deferred", "reason": "已读CASK method/experimentaldesign：DAS rank16 trainonlyownershipcausalintervention，relation-invarianttypedexport保world/branch/speaker作用域，不存脆弱坐标。下一批需真实activation和commitboundary，frozenmemory原文不是本地事实，provisional不能升authority。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "4 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experimental design", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09008v1", "sha256": "18b66f7b6a517c4c1ed1b7cf728f222afc5de4d86be7c68610940623807f63bc", "local_html_sha256": "18b66f7b6a517c4c1ed1b7cf728f222afc5de4d86be7c68610940623807f63bc", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09008v1", "result": "retrieved", "sha256": "18b66f7b6a517c4c1ed1b7cf728f222afc5de4d86be7c68610940623807f63bc" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.09003", "title": "Algorithmic Scratchpads and Curriculum Staging for Arithmetic Reasoning in Tiny Transformers", "published": "2026-10-06T18:58:13Z", "reviews": [ { "id": "2610.09003", "title": "Algorithmic Scratchpads and Curriculum Staging for Arithmetic Reasoning in Tiny Transformers", "published": "2026-10-06T18:58:13Z", "decision": "reviewed_deferred", "reason": "TinyTransformer scratchpad/curriculum与packing修gradientstarvation，4000heldout arithmetic；四digit0%/遗忘0%边界，不能夸为普遍算法泛化。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Experimental Setup and Architectures", "location": ".cache/oct12-scan/fulltext/2610.09003.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "3 Experimental Setup and Architectures", "location": ".cache/oct12-scan/fulltext/2610.09003.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" }, { "role": "additional targeted method/evaluation", "heading": "7 Large-Scale 4,000-Problem Benchmark and Developmental Dynamics", "section_index": 6, "characters_read": [ 0, 1500 ], "scope": "bounded excerpt; original source SHA verified" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 244, "primary_fulltext": { "url": "https://arxiv.org/html/2610.09003v1", "sha256": "ed238b305e681cc26521e1d3413765ed0f74a152e94676f1dfa5fcee31a80cc1", "local_path": ".cache/oct12-scan/fulltext/2610.09003.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09003v1", "result": "retrieved", "sha256": "ed238b305e681cc26521e1d3413765ed0f74a152e94676f1dfa5fcee31a80cc1" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.09003", "sha256": "5433b69b81aafabb191134fb826432d1b71467b1136e52f6fe37c791dd2f8a0a", "local_path": ".cache/oct12-scan/fulltext/2610.09003.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Tue, 6 Oct 2026 18:58:13 UTC (118 KB)", "published": "2026-10-06T18:58:13Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.09000", "title": "How Fragile Is On-Device Language Model Safety? Localizing Safety-Critical Parameters for Sparse Fault Analysis", "published": "2026-10-06T18:56:14Z", "reviews": [ { "id": "2610.09000", "title": "How Fragile Is On-Device Language Model Safety? Localizing Safety-Critical Parameters for Sparse Fault Analysis", "published": "2026-10-06T18:56:14Z", "decision": "reviewed_deferred", "reason": "已读safetyweights method/experiment：Llama2单模型lowranksubspace与parameterimportance、420attribution/100AdvBench eval分离。foundationsecurity候选待下一批严格sandbox，不能将稀疏safetyremoval当agent能力改进或按test选安全参数。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Methodology", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experimental Setup", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.09000v1", "sha256": "4b2bcefde52574b1468cf592472bec00a9098426a1ffdade8e9df2c8264764c0", "local_html_sha256": "4b2bcefde52574b1468cf592472bec00a9098426a1ffdade8e9df2c8264764c0", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.09000v1", "result": "retrieved", "sha256": "4b2bcefde52574b1468cf592472bec00a9098426a1ffdade8e9df2c8264764c0" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.10592", "title": "Wieszcz-XIX: A 3.1-Billion-Word Corpus of Pre-1918 Polish and Temporally Bounded Language Models Trained From Scratch", "published": "2026-10-06T18:35:26Z", "reviews": [ { "id": "2610.10592", "title": "Wieszcz-XIX: A 3.1-Billion-Word Corpus of Pre-1918 Polish and Temporally Bounded Language Models Trained From Scratch", "published": "2026-10-06T18:35:26Z", "decision": "reviewed_deferred", "reason": "WieszczXIX historicalPolishcorpus documentdedup/时间泄漏审计与47-349Mfromscratchmodel ladder，denseheldouttemporalBPB；foundationpretrain候选nextbatch，OCR45%不可纠正和残余latertext须保留，非posttrainingRL。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Corpus construction", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "7 Evaluation", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.10592v1", "sha256": "6dfce9a2d889083032063994c46b556b89141c6ea204c47e11b1f95e7f7b77db", "local_html_sha256": "6dfce9a2d889083032063994c46b556b89141c6ea204c47e11b1f95e7f7b77db", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.10592v1", "result": "retrieved", "sha256": "6dfce9a2d889083032063994c46b556b89141c6ea204c47e11b1f95e7f7b77db" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.08963", "title": "On KL-Regularized Policy Optimization", "published": "2026-10-06T18:30:42Z", "reviews": [ { "id": "2610.08963", "title": "On KL-Regularized Policy Optimization", "published": "2026-10-06T18:30:42Z", "decision": "reviewed_deferred", "reason": "官方PDF已取得并限定阅读方法/结论。KLPO以sampler为KL锚点，将Gibbs最优性转为log-ratio最小二乘；profile intercept需要真实sampler-to-trainer KL，独立MC-KL/交叉拟合防止偏差，不能以同一动作或截断KL冒充无偏更新。结论明确大规模实证为后续工作，保留理论/后训练机制下一批，非现成已验证LLM能力收益。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_official_pdf_method_and_evaluation_pages", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "Title/author and arXiv v1 margin", "pdf_page": 1, "scope": "full page text read; manuscript cover date is not substituted for arXiv v1 date" }, { "heading": "Method and evaluation/conclusion", "pdf_pages": [ 5, 6, 8, 9, 11, 13, 14, 15 ], "scope": "listed page text read; not all article pages or appendices", "method_evidence": "sampler version-specific probabilities/conditionals；Gibbs log-ratio residual；profiled intercept = sampler signal mean + sampler-to-trainer KL；critic-free token PMD终端return；独立辅助抽样MC-KL；序列unbiased estimator需要crossfit/U-statistic至少两个辅助draw，保留完整非过滤rollout且token求和、response求均值。", "evaluation_evidence": "PDF第15页结论将large-scale empirical evaluation、finite-sample/convergence analysis列为next steps；本次相关页没有可用于宣称本地或原文LLM benchmark增益的实验数值。未通读53页所有附录，不作整篇零实验断言。" } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.08963v1", "sha256": null, "local_html_sha256": null, "local_html_sha256_matches": null, "attempts": [ { "url": "https://arxiv.org/html/2610.08963v1", "result": "failed", "error": "404 Client Error: Not Found for url: https://arxiv.org/html/2610.08963v1" }, { "url": "https://arxiv.org/html/2610.08963v1", "result": "failed", "error": "404 Client Error: Not Found for url: https://arxiv.org/html/2610.08963v1" }, { "url": "https://arxiv.org/pdf/2610.08963v1", "result": "retrieved", "retrieved_on": "2026-10-12", "command": "curl -sSL --fail --max-time 45 https://arxiv.org/pdf/2610.08963v1 -o tmp/pdfs/2610.08963.pdf", "sha256": "de3e299d401af2715a49090328f86a8e4ba284983e189d8119912f4a573f4590" } ], "alternate_evidence": { "url": "https://arxiv.org/pdf/2610.08963v1", "format": "official_arxiv_pdf", "sha256": "de3e299d401af2715a49090328f86a8e4ba284983e189d8119912f4a573f4590", "local_path": "tmp/pdfs/2610.08963.pdf", "local_sha256_verified": true, "total_pdf_pages": 53 } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.08959", "title": "GraphOPD: Graph-Augmented On-Policy Distillation for LLM Agents", "published": "2026-10-06T18:25:23Z", "reviews": [ { "id": "2610.08959", "title": "GraphOPD: Graph-Augmented On-Policy Distillation for LLM Agents", "published": "2026-10-06T18:25:23Z", "decision": "reviewed_deferred", "reason": "GraphOPD用environment entity Footprints构dependency graph，stationary random walk结构credit与divergence融合轨迹相对mask；ALFWorld/WebShop/SearchQA有共享GRPO与executed-replay audit，下一批必须真实footprints/因果回放而非gold plan。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/post-training-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "4 Experiments", "scope": "bounded section excerpt, not the full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.08959v1", "sha256": "5395b73753261a0ae4f9f6ad90d1788fc0f2fa114d8054f5de172934ff6a2619", "local_html_sha256": "5395b73753261a0ae4f9f6ad90d1788fc0f2fa114d8054f5de172934ff6a2619", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.08959v1", "result": "retrieved", "sha256": "5395b73753261a0ae4f9f6ad90d1788fc0f2fa114d8054f5de172934ff6a2619" } ], "alternate_evidence": null }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "post-training" ], "status": "reviewed_deferred" }, { "id": "2610.08954", "title": "RACER: Reflective Agent Coupling Query Interpretation and Tool-Based Retrieval for Frame Selection in Long Video Understanding", "published": "2026-10-06T18:21:13Z", "reviews": [ { "id": "2610.08954", "title": "RACER: Reflective Agent Coupling Query Interpretation and Tool-Based Retrieval for Frame Selection in Long Video Understanding", "published": "2026-10-06T18:21:13Z", "decision": "reviewed_deferred", "reason": "轻量VidLLM把complexquery改为M=3 evidence-subqueries，CLIP定位真实video frames，反馈frames反思query再检索；query模型不直接决定frames。; Video-MME无字幕、MLVU、LongVideoBench，四answeringVLM32-frame budget；3Bqueryinterpreter+ViT-L/14工具。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract + explicitly bounded original method/evaluation passages (ranges recorded); not complete main sections or appendices", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.08954v1#S3", "https://arxiv.org/html/2610.08954v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.08954v1", "sha256": "6563e3ee1075a0019203237bb8d0d87bf1dd4c0fa62c2289dd16582a24eacb33" }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent", "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.08941", "title": "SPW-Nav: A Streaming Panoramic World Model for Language-Guided Navigation", "published": "2026-10-06T18:05:45Z", "reviews": [ { "id": "2610.08941", "title": "SPW-Nav: A Streaming Panoramic World Model for Language-Guided Navigation", "published": "2026-10-06T18:05:45Z", "decision": "reviewed_deferred", "reason": "SPWNav sphericalrotationexact/posealignedtranslation/memory与Wan5B fewstep，分钟panorama实时；datasetlanguageverified与switchcontrol真worldmodel待核。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "4 Methodology", "location": ".cache/oct12-scan/fulltext/2610.08941.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "5 Experiments", "location": ".cache/oct12-scan/fulltext/2610.08941.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 245, "primary_fulltext": { "url": "https://arxiv.org/html/2610.08941v1", "sha256": "3b00c35c09a769b9f5ecdc458a6752b0b3ca932355c5cc2ffadf87d7f26bcf72", "local_path": ".cache/oct12-scan/fulltext/2610.08941.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.08941v1", "result": "retrieved", "sha256": "3b00c35c09a769b9f5ecdc458a6752b0b3ca932355c5cc2ffadf87d7f26bcf72" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.08941", "sha256": "52c624c17c9c4188cb504a9d612e0d759100ecf44c5d6eec2cb5daf0cae1153e", "local_path": ".cache/oct12-scan/fulltext/2610.08941.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Tue, 6 Oct 2026 18:05:45 UTC (36,690 KB)", "published": "2026-10-06T18:05:45Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.08928", "title": "Multi-Aspect Runtime Verification for Simulation-Based V&V of LLM-Enabled Autonomous Agents", "published": "2026-10-06T18:00:43Z", "reviews": [ { "id": "2610.08928", "title": "Multi-Aspect Runtime Verification for Simulation-Based V&V of LLM-Enabled Autonomous Agents", "published": "2026-10-06T18:00:43Z", "decision": "reviewed_deferred", "reason": "已读multiaspect runtimeframework/results高层：typed spatial/temporal/semantic、canonicalevent与4valueprovenance、MonPoly/reference逐时一致，simulation safetyblocking。下一批可核民用安全监控机制，受控九attack及noobservedFP不等现实安全保证，不运行现实防务系统。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Multi-Aspect Runtime Assurance Framework", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experimental Results", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.08928v1", "sha256": "7d83d16ca6af3c2491919ecff7c9cc8954a6605e071964ab3092a4cefa0c6f81", "local_html_sha256": "7d83d16ca6af3c2491919ecff7c9cc8954a6605e071964ab3092a4cefa0c6f81", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.08928v1", "result": "retrieved", "sha256": "7d83d16ca6af3c2491919ecff7c9cc8954a6605e071964ab3092a4cefa0c6f81" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.08917", "title": "CARE: Certifying Acceleration for Vision-Language-Action Inference", "published": "2026-10-06T18:00:05Z", "reviews": [ { "id": "2610.08917", "title": "CARE: Certifying Acceleration for Vision-Language-Action Inference", "published": "2026-10-06T18:00:05Z", "decision": "reviewed_deferred", "reason": "paired episode AIF=reference success且candidate fail；独立profiling排序、task-balanced round、family correction的anytime e-process、failure-triggered reference cache、无certificate回退reference。; §4 LIBERO四suite与Crafter、独立calibration/test、上万resampling；包括24候选、flowsteps、不同模型的closed-loop完整rollouts。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract/author information/main defining methods/main experiments and tables; production-evidence paragraph spot check; appendices not exhaustively read", "review_artifact": "research-audits/oct12-review-closure/foundation-model-reviews.json", "evidence_locations": [ "https://arxiv.org/html/2610.08917v1#S3", "https://arxiv.org/html/2610.08917v1#S4" ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.08917v1", "sha256": "304d8813c0603a87a2faba6aed20023384930934c40a8132e0d770fe2d627291" }, "new_implementation_inferred_from_review": false }, { "id": "2610.08917", "title": "CARE: Certifying Acceleration for Vision-Language-Action Inference", "published": "2026-10-06T18:00:05Z", "decision": "reviewed_deferred", "reason": "Google合作CARE wholeepisodepairedAIF、sequentialcertificate/familycorrection与referencefallback；LIBERO/Crafter非productionAB，真latency/pairedrollout不可手造。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_listed_bounded_primary_method_and_evaluation_excerpts", "review_artifact": "research-audits/oct12-review-closure/foundation-tail-reviews.json", "evidence_locations": [ { "role": "method", "heading": "3 Approach", "location": ".cache/oct12-scan/fulltext/2610.08917.json sections text prefix", "characters_read": [ 0, 600 ], "scope": "bounded excerpt, not entire section" }, { "role": "evaluation", "heading": "4 Experiments", "location": ".cache/oct12-scan/fulltext/2610.08917.json sections text prefix", "characters_read": [ 0, 750 ], "scope": "bounded excerpt, not entire section" } ], "dateproof": null, "source_proof": { "abstract_assignment_path": ".cache/oct12-scan/fulltext/foundation-model-assignment.json", "abstract_assignment_sha256": "9ba93117ba8a6f8d708b9b39ed424d8dee24e67fff5e17daabc2f62138e5f87d", "abstract_assignment_index": 246, "primary_fulltext": { "url": "https://arxiv.org/html/2610.08917v1", "sha256": "304d8813c0603a87a2faba6aed20023384930934c40a8132e0d770fe2d627291", "local_path": ".cache/oct12-scan/fulltext/2610.08917.html", "local_sha256_matches_recorded": true, "attempts": [ { "url": "https://arxiv.org/html/2610.08917v1", "result": "retrieved", "sha256": "304d8813c0603a87a2faba6aed20023384930934c40a8132e0d770fe2d627291" } ] }, "primary_date": { "url": "https://arxiv.org/abs/2610.08917", "sha256": "f4dbefe13bf80bd41165fa980264d55071b66b1976c57a4d39e1b325e9f87f55", "local_path": ".cache/oct12-scan/fulltext/2610.08917.abs.html", "locator": "Submission history [v1], first dated entry", "raw_v1_timestamp": "Tue, 6 Oct 2026 18:00:05 UTC (2,105 KB)", "published": "2026-10-06T18:00:05Z", "assignment_timestamp_matches": true, "date_kind": "arxiv_v1_not_claimed_earliest_public_release" } }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "foundation-model" ], "status": "reviewed_deferred" }, { "id": "2610.08791", "title": "World Models' Last Exam in Physics", "published": "2026-10-06T17:59:56Z", "reviews": [ { "track": "foundation-model", "id": "2610.08791", "title": "World Models' Last Exam in Physics", "published": "2026-10-06T17:59:56Z", "repository_status": "new", "pool": "main", "decision": "accepted", "reason": "40项物理关系测量任务覆盖九类现象，官方公开任务图像/提示/评估器及外部视频批评估入口，可纳入evolve公共评测候选；不是新世界模型算法，也不是已接入。", "review_level": "fulltext", "kind": "evaluation_infrastructure", "official_url": "https://arxiv.org/html/2610.08791v1", "code_url": "https://github.com/Einsia/phys-last-exam", "locations": [ "§2.1–2.4/Eq1–7", "§3.1–3.3/Table2", "official README", "docs/evaluation.md", "easy/P21/evaluator/_shared/unified_evaluators/runtime.py:339–355", "evaluate.py:summarize" ], "evidence": [ { "url": "https://arxiv.org/abs/2610.08791", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08791v1", "location": "§2.1–2.4/Eq1–7; §3.1–3.3/Table2; official README; docs/evaluation.md; easy/P21/evaluator/_shared/unified_evaluators/runtime.py:339–355; evaluate.py:summarize" }, { "url": "https://arxiv.org/html/2610.08791v1", "location": "§2.1–2.4/Eq1–7; §3.1–3.3/Table2; official README; docs/evaluation.md; easy/P21/evaluator/_shared/unified_evaluators/runtime.py:339–355; evaluate.py:summarize", "notes": "预设物理残差评分，VLM仅负责可观测性门；40task×8model×4sample，human排名51.58%对43.40% VLM，pairwise53.12%对45.62%。" } ], "limitations": "官方网页404但GitHub可访问。已确认论文与代码差异：论文门失败仍尝试物理测量，公开runtime gate失败直接return/physics_attempted=False；接入须保留原论文测量协议并分别报告coverage，不可把现有脚本直接宣称保真实现。难度标签基于被评模型均分事后划分；human一致性仅约53%，不是物理真值oracle。", "implementation_status": "not_implemented", "abstract_review_completed": true, "fulltext_review_completed": true, "fulltext_notes": "预设物理残差评分，VLM仅负责可观测性门；40task×8model×4sample，human排名51.58%对43.40% VLM，pairwise53.12%对45.62%。", "review_status": "reviewed" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08779", "title": "ALIVE: Interaction-Aligned Object Insertion for First-Frame-Guided Video Editing", "published": "2026-10-06T17:58:51Z", "reviews": [ { "track": "foundation-model", "id": "2610.08779", "title": "ALIVE: Interaction-Aligned Object Insertion for First-Frame-Guided Video Editing", "published": "2026-10-06T17:58:51Z", "repository_status": "new", "pool": "main", "decision": "accepted", "reason": "ALIVE以source+noisy target通道拼接、干净首帧和按时间overlap/描述长度归一的cross-attention；Qwen VLM从同输入预测P4而不是读target。35,800对训练与两benchmark；部分基线额外P2提示不完全同条件。", "review_level": "fulltext", "abstract_review_completed": true, "fulltext_review_completed": true, "evidence": [ { "url": "https://arxiv.org/abs/2610.08779", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08779v1", "location": "§3；§4.2式2–4、§4.3式5；§5.1" }, { "url": "https://arxiv.org/html/2610.08779v1", "location": "§3；§4.2式2–4、§4.3式5；§5.1", "notes": "ALIVE以source+noisy target通道拼接、干净首帧和按时间overlap/描述长度归一的cross-attention；Qwen VLM从同输入预测P4而不是读target。35,800对训练与两benchmark；部分基线额外P2提示不完全同条件。" } ], "review_status": "reviewed", "official_url": "https://arxiv.org/html/2610.08779v1", "locations": [ "§3；§4.2式2–4、§4.3式5；§5.1" ], "implementation_status": "not_implemented", "fulltext_notes": "ALIVE以source+noisy target通道拼接、干净首帧和按时间overlap/描述长度归一的cross-attention；Qwen VLM从同输入预测P4而不是读target。35,800对训练与两benchmark；部分基线额外P2提示不完全同条件。" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08901", "title": "Sequential Probabilistic Uncertainty Estimation for Parallel Multi-Agent Reasoning Systems", "published": "2026-10-06T17:58:43Z", "reviews": [ { "id": "2610.08901", "title": "Sequential Probabilistic Uncertainty Estimation for Parallel Multi-Agent Reasoning Systems", "published": "2026-10-06T17:58:43Z", "decision": "reviewed_deferred", "reason": "已读SAUCE method/experiment：真实parallelMAS round agreement与generationuncertainty的filtering递归，misclassification/selectiveprediction及PlattcalibratedBrier。下一批需要真实LMprobability与validation-onlycalibration，不能把testcorrectness喂不确定性估计器。", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "read_scope": "abstract_plus_bounded_primary_section_excerpts", "review_artifact": "research-audits/oct12-review-closure/agent-reviews.json", "evidence_locations": [ { "heading": "3 Method", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] }, { "heading": "5 Experiments", "scope": "bounded section excerpt, not full section", "characters_read_at_least": [ 0, 650 ] } ], "dateproof": null, "source_proof": { "url": "https://arxiv.org/html/2610.08901v1", "sha256": "22f9854744e4b39366ad9a261bfad4badeb996320110ea85f9701ac0ebdc6fdc", "local_html_sha256": "22f9854744e4b39366ad9a261bfad4badeb996320110ea85f9701ac0ebdc6fdc", "local_html_sha256_matches": true, "attempts": [ { "url": "https://arxiv.org/html/2610.08901v1", "result": "retrieved", "sha256": "22f9854744e4b39366ad9a261bfad4badeb996320110ea85f9701ac0ebdc6fdc" } ] }, "new_implementation_inferred_from_review": false } ], "implementation_status": "not_implemented_by_review", "discovery_origin": "configured-arxiv-recall", "tracks": [ "agent" ], "status": "reviewed_deferred" }, { "id": "2610.08775", "title": "Agent in a Bottle: Can LLM Agents Turn Their Capabilities Into Cheap, Scalable Artifacts?", "published": "2026-10-06T17:57:19Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08775", "title": "Agent in a Bottle: Can LLM Agents Turn Their Capabilities Into Cheap, Scalable Artifacts?", "published": "2026-10-06T17:57:19Z", "source_group": "original", "review_level": "fulltext", "source": "https://arxiv.org/html/2610.08775v1", "location": "§3.1–3.5 Eq1–6；§4.1；Ethics/Reproducibility；官方README与src/bottled/datasets/esci.py", "decision": "accepted", "reason": "BOTTLED作为evolve可执行评测基础设施接受：Agent在预算内生成可复用artifact并完成大批无标签任务，明确隐藏gold与预算约束，兼容模型/脚本/路由器候选。", "reviewer": "agent_review", "evidence": "MAVE/ESCI/RAID三任务；5M加权tokens、10h、A10040G；官方代码存在agent_eval.py/预算MCP/proxy/数据与评分器；ESCI源码拆questions/gold并确定fallback后计算macroF1。", "limitations": "原实验每模型每任务两runs，零样本只抽1000；GPU成本未统一入token预算；MIT代码，MAVE CC-BY-NC4.0/ESCI Apache2.0/RAID MIT；contamination judge及Jev评测未公开，不能宣称完整污染审计复现。", "code": "https://github.com/aktsonthalia/bottled", "license": "MIT code; dataset-specific licenses", "candidate_kind": "evaluation_infrastructure", "fulltext_review_completed": true, "review_status": "reviewed", "implementation_status": "not_implemented_in_this_review" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08720", "title": "WorldSolver: Can LLM Agents Simulate the Physical Dynamics via Solver Generation?", "published": "2026-10-06T17:26:26Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08720", "title": "WorldSolver: Can LLM Agents Simulate the Physical Dynamics via Solver Generation?", "published": "2026-10-06T17:26:26Z", "source_group": "original", "review_level": "fulltext", "source": "https://arxiv.org/html/2610.08720v1", "location": "§2.1–2.3 Eq1–10；§3.1–3.3/Table3；Limitations；官方仓库/API访问2026-10-07", "decision": "pending", "reason": "WorldSolver的168物理求解器任务/固定scaffold/执行+VLM视觉+物理规则评分适合evolve，但官方链接github.com/sirujiang/WorldSolver网页与API均404，暂不能确认公开任务/评分器/许可，等待可执行发布。", "reviewer": "agent_review", "evidence": "7物理域；生成1800秒/solver600秒；执行失败记0，成功取(V+P)/2；GPT5.6Sol48.7、ClaudeOpus546.7。", "limitations": "60分pass阈值仅演示非验证标准；评分规则对Agent隐藏；不同模型用不同harness；VLM judge也参与被测模型，需独立审核；代码不可达不是算法无价值。", "code": "https://github.com/sirujiang/WorldSolver", "pending_reason": "public_artifacts_unavailable", "candidate_kind": "evaluation_infrastructure", "review_status": "reviewed_deferred", "fulltext_review_completed": true, "implementation_status": "not_implemented_in_this_review" } ], "status": "reviewed_deferred", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08718", "title": "When Forgetting is not Catastrophic: On the Mechanics of Spurious Forgetting", "published": "2026-10-06T17:25:48Z", "reviews": [ { "track": "foundation-model", "id": "2610.08718", "title": "When Forgetting is not Catastrophic: On the Mechanics of Spurious Forgetting", "published": "2026-10-06T17:25:48Z", "repository_status": "new", "pool": "main", "decision": "accepted", "reason": "新事实微调造成共享表征偏移与逐事实侵蚀分离；对更新矩阵删首奇异分量的干预在OLMo2-1B有同范数随机rank1对照，可登记机制诊断候选，不能宣传为通用抗遗忘能力算法。", "review_level": "fulltext", "abstract_review_completed": true, "fulltext_review_completed": true, "evidence": [ { "url": "https://arxiv.org/abs/2610.08718", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08718v1", "location": "§4.1–4.4; §5.1–5.2/Figure5; §6/Limitations; Appendix D.2" }, { "url": "https://arxiv.org/html/2610.08718v1", "location": "§4.1–4.4; §5.1–5.2/Figure5; §6/Limitations; Appendix D.2", "notes": "三seed、CounterFact初始已知旧事实；合成新实体训练恢复但真实EntityQuestions制造商不恢复；删主方向可还原旧事实但损新事实，随机方向不能。" } ], "kind": "mechanism_diagnostic", "official_url": "https://arxiv.org/html/2610.08718v1", "locations": [ "§4.1–4.4", "§5.1–5.2/Figure5", "§6/Limitations", "Appendix D.2" ], "limitations": "人为新旧答案首token不重叠，单预训练模型；诊断logit均值来自已知oldfact不能在线oracle使用。只采用权重干预时须独立validation选checkpoint和rank。", "implementation_status": "not_implemented", "fulltext_notes": "三seed、CounterFact初始已知旧事实；合成新实体训练恢复但真实EntityQuestions制造商不恢复；删主方向可还原旧事实但损新事实，随机方向不能。", "review_status": "reviewed" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08713", "title": "SpaTime: Streaming Vision-Language Models for Spatio-temporal Reasoning", "published": "2026-10-06T17:22:33Z", "reviews": [ { "track": "foundation-model", "id": "2610.08713", "title": "SpaTime: Streaming Vision-Language Models for Spatio-temporal Reasoning", "published": "2026-10-06T17:22:33Z", "repository_status": "new", "pool": "main", "decision": "accepted", "reason": "SpaTime冻结因果几何/视觉编码器，2×2 merge投影相加，训练LoRA/state/head并建模首个Response分布作时间损失；strict只计最早回答，charitable末帧强制回答须分开。StreamVSTI训练、StreamVSI任务OOD。", "review_level": "fulltext", "abstract_review_completed": true, "fulltext_review_completed": true, "evidence": [ { "url": "https://arxiv.org/abs/2610.08713", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08713v1", "location": "§4.1式5；§4.2式6–8；§5；§6" }, { "url": "https://arxiv.org/html/2610.08713v1", "location": "§4.1式5；§4.2式6–8；§5；§6", "notes": "SpaTime冻结因果几何/视觉编码器，2×2 merge投影相加，训练LoRA/state/head并建模首个Response分布作时间损失；strict只计最早回答，charitable末帧强制回答须分开。StreamVSTI训练、StreamVSI任务OOD。" } ], "review_status": "reviewed", "official_url": "https://arxiv.org/html/2610.08713v1", "locations": [ "§4.1式5；§4.2式6–8；§5；§6" ], "implementation_status": "not_implemented", "fulltext_notes": "SpaTime冻结因果几何/视觉编码器，2×2 merge投影相加，训练LoRA/state/head并建模首个Response分布作时间损失；strict只计最早回答，charitable末帧强制回答须分开。StreamVSTI训练、StreamVSI任务OOD。" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08691", "title": "ScienceClaw: Benchmarking Continual Self-Evolution of AI-for-Science Agents Across the Natural and Social Sciences", "published": "2026-10-06T17:08:03Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08691", "title": "ScienceClaw: Benchmarking Continual Self-Evolution of AI-for-Science Agents Across the Natural and Social Sciences", "published": "2026-10-06T17:08:03Z", "source_group": "original", "review_level": "fulltext", "source": "https://arxiv.org/html/2610.08691v1", "location": "§3 Eq1–3；§4.1–4.2；§5.1–5.2 Eq5–13；§6 Tables1–2/Fig6；§8", "decision": "accepted", "reason": "ScienceClaw冻结模型，以typed workflow执行修复提取linked Skill–Operator bundles，boundary/source reset replay必须实际使用新版本，只有独立validation严格提升且hard constraints/预算通过才原子晋级；符合真实evolve闭环候选。", "reviewer": "agent_review", "evidence": "23学科每学科64 IID+64 OOD；7轮MacroSR OOD77.72→91.30；70/161候选晋级；OOD不参与提案/选择；去链接/去replay/独立validator均有消融。", "limitations": "主要final snapshots，作者§8承认广泛结论需重复runs和学科intervals；domain evaluator可能漏科学错误；需实际科学工具/独立verifier，不能固定轨迹fixture作为能力复现。", "code": "https://github.com/beita6969/ScienceClaw", "fulltext_review_completed": true, "review_status": "reviewed", "implementation_status": "not_implemented_in_this_review" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08678", "title": "Secure Speculative Decoding for Large Language Models", "published": "2026-10-06T16:59:37Z", "reviews": [ { "track": "foundation-model", "id": "2610.08678", "title": "Secure Speculative Decoding for Large Language Models", "published": "2026-10-06T16:59:37Z", "repository_status": "new", "pool": "main", "decision": "accepted", "reason": "SecureSD 按全局生成位置混合标准/宽松接受与拒绝恢复分布，早期严格验证，后续恢复快速验证；定义完整，可作为真实投机解码候选。", "review_level": "fulltext", "abstract_review_completed": true, "fulltext_review_completed": true, "evidence": [ { "url": "https://arxiv.org/abs/2610.08678", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08678v1", "location": "§IV-B; §V Algorithm 2 / Eq7–14; §VII-A paired protocol; §VIII Tables XII–XIII" }, { "url": "https://arxiv.org/html/2610.08678v1", "location": "§IV-B; §V Algorithm 2 / Eq7–14; §VII-A paired protocol; §VIII Tables XII–XIII", "notes": "" } ], "kind": "inference_algorithm", "locations": [ "§IV-B", "§V Algorithm 2 / Eq7–14", "§VII-A paired protocol", "§VIII Tables XII–XIII" ], "limitations": "不是对任意攻击的安全保证；需目标/草稿模型、lossless/relaxed/corrected 三臂同参数对照，并把任务质量和真实延迟分开报告。原文主要 Blackwell 实测，不代表 A100 已验证。", "code_url": "https://github.com/YichiCS/Secure-Speculative-Decoding", "code_status": "official repository accessible; MIT; bench/securesd/scripts present; some judge/model checkpoints gated", "official_url": "https://arxiv.org/html/2610.08678v1", "review_provenance": "root 独立正文审查；foundation读取整合", "implementation_status": "not_implemented", "fulltext_notes": "", "review_status": "reviewed" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08674", "title": "EC-RAG: Event Chain Retrieval-Augmented Generation for Long Video Understanding", "published": "2026-10-06T16:57:46Z", "reviews": [ { "track": "foundation-model", "id": "2610.08674", "title": "EC-RAG: Event Chain Retrieval-Augmented Generation for Long Video Understanding", "published": "2026-10-06T16:57:46Z", "repository_status": "new", "pool": "main", "decision": "accepted", "reason": "EC-RAG训练外query模态拆解、CLIP variance取帧、ASR/OCR/DET事件描述与按时间相邻链，Contriever相关事件/邻接上下文辅助回答；三长视频benchmark，不是学习因果事件图，额外工具/token预算须计。", "review_level": "fulltext", "abstract_review_completed": true, "fulltext_review_completed": true, "evidence": [ { "url": "https://arxiv.org/abs/2610.08674", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08674v1", "location": "§3.1–3.4式12–16；§4.1–4.2" }, { "url": "https://arxiv.org/html/2610.08674v1", "location": "§3.1–3.4式12–16；§4.1–4.2", "notes": "EC-RAG训练外query模态拆解、CLIP variance取帧、ASR/OCR/DET事件描述与按时间相邻链，Contriever相关事件/邻接上下文辅助回答；三长视频benchmark，不是学习因果事件图，额外工具/token预算须计。" } ], "review_status": "reviewed", "official_url": "https://arxiv.org/html/2610.08674v1", "locations": [ "§3.1–3.4式12–16；§4.1–4.2" ], "implementation_status": "not_implemented", "fulltext_notes": "EC-RAG训练外query模态拆解、CLIP variance取帧、ASR/OCR/DET事件描述与按时间相邻链，Contriever相关事件/邻接上下文辅助回答；三长视频benchmark，不是学习因果事件图，额外工具/token预算须计。" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08668", "title": "Semantic Behavioral Watermarking: Paraphrase-Robust and Forgery-Resistant Provenance for LLM Agents", "published": "2026-10-06T16:50:42Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08668", "title": "Semantic Behavioral Watermarking: Paraphrase-Robust and Forgery-Resistant Provenance for LLM Agents", "published": "2026-10-06T16:50:42Z", "source_group": "original", "review_level": "fulltext", "source": "https://arxiv.org/html/2610.08668v1", "location": "§3 threat model；§4.1–4.3；§5；§6.1/6.3/6.4 Tables2/4/5；§7", "decision": "out_of_scope", "reason": "SBW的语义聚类/历史keyed PRF/弃权margin为Agent行为溯源水印算法，主要优化归因检测而非Agent规划、学习、工具使用或任务能力，排除本轮核心能力收录。", "reviewer": "agent_review", "evidence": "ToolBench paraphrase TPR 0.49–0.66；ALFWorld 0.92–0.97；训练无关encoder+kmeans+permutation detector。", "limitations": "§7明示ToolBench无环境执行反馈；复制/链式replay未解决；utility只是动作与argmax一致率，不是任务成功率；部分鲁棒审计artifacts未保留。", "fulltext_review_completed": true, "review_status": "reviewed", "implementation_status": "not_implemented_in_this_review" } ], "status": "rejected_or_out_of_scope", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08662", "title": "ParanoiaEval: Benchmarking Unnecessary Defensive Work in Agentic Coding", "published": "2026-10-06T16:46:14Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08662", "title": "ParanoiaEval: Benchmarking Unnecessary Defensive Work in Agentic Coding", "published": "2026-10-06T16:46:14Z", "source_group": "original", "review_level": "fulltext", "source": "https://arxiv.org/html/2610.08662v1", "location": "§4.1–4.3；§5；§7 Code/Data；官方README、scripts/judge.py", "decision": "accepted", "reason": "ParanoiaEval作为evolve行为约束评测接受：200同任务E+/E−证据对，任务oracle和多余风险处置judge分离，可同时评测完成度与证据响应。", "reviewer": "agent_review", "evidence": "50固定repo、400variants；oracle检验baseline失败/reference成功（28read-only例外）；冻结64人标memory、跨task检索；heldout200判决accuracy.965/kappa.93；MIT任务/runner/judge公开。", "limitations": "公开包无结果/完整轨迹；answer cards/reference补丁只能独立scorer访问；模型native harness混杂不能声称模型净能力差异；developer study限成功runs且非部署A/B；默认same-task=none，不能打开other-arm泄漏gold。", "code": "https://github.com/ZhuoningXu/ParanoiaEval_release", "license": "MIT", "candidate_kind": "evaluation_infrastructure", "fulltext_review_completed": true, "review_status": "reviewed", "implementation_status": "not_implemented_in_this_review" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08659", "title": "Selective Transfer of RL Updates for Visual Reasoning", "published": "2026-10-06T16:43:55Z", "reviews": [ { "track": "foundation-model", "id": "2610.08659", "title": "Selective Transfer of RL Updates for Visual Reasoning", "published": "2026-10-06T16:43:55Z", "repository_status": "new", "pool": "main", "decision": "accepted", "reason": "Selective-RL隔离R−B与B−V，逐矩阵rank1 SVD后恢复原Frobenius norm，分别重构donor插值和直接加RLdelta；3个8B VLM骨干，源验证选donor，H1验证选λ后冻结，需保留非视觉模块支持对齐。", "review_level": "fulltext", "abstract_review_completed": true, "fulltext_review_completed": true, "evidence": [ { "url": "https://arxiv.org/abs/2610.08659", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08659v1", "location": "§3.1–3.3式1–3；§4.1；Appendix A Algorithm 1" }, { "url": "https://arxiv.org/html/2610.08659v1", "location": "§3.1–3.3式1–3；§4.1；Appendix A Algorithm 1", "notes": "Selective-RL隔离R−B与B−V，逐矩阵rank1 SVD后恢复原Frobenius norm，分别重构donor插值和直接加RLdelta；3个8B VLM骨干，源验证选donor，H1验证选λ后冻结，需保留非视觉模块支持对齐。" } ], "review_status": "reviewed", "official_url": "https://arxiv.org/html/2610.08659v1", "locations": [ "§3.1–3.3式1–3；§4.1；Appendix A Algorithm 1" ], "implementation_status": "not_implemented", "fulltext_notes": "Selective-RL隔离R−B与B−V，逐矩阵rank1 SVD后恢复原Frobenius norm，分别重构donor插值和直接加RLdelta；3个8B VLM骨干，源验证选donor，H1验证选λ后冻结，需保留非视觉模块支持对齐。" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08651", "title": "A Case Study in Assuring AI-Written Software", "published": "2026-10-06T16:37:08Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08651", "title": "A Case Study in Assuring AI-Written Software", "published": "2026-10-06T16:37:08Z", "source_group": "original", "review_level": "abstract", "source": "https://arxiv.org/abs/2610.08651", "location": "Title and complete abstract", "decision": "out_of_scope", "reason": "医疗软件治理案例研究，无独立Agent算法贡献。", "implementation_status": "not_implemented_in_this_review", "review_status": "reviewed" } ], "status": "rejected_or_out_of_scope", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08649", "title": "Stable Scores, Unstable Answers: Frame Phase and Option Order in Video Multiple-Choice Evaluation", "published": "2026-10-06T16:36:06Z", "reviews": [ { "track": "foundation-model", "id": "2610.08649", "title": "Stable Scores, Unstable Answers: Frame Phase and Option Order in Video Multiple-Choice Evaluation", "published": "2026-10-06T16:36:06Z", "repository_status": "new", "pool": "main", "decision": "accepted", "reason": "PhaseFusion把3×8帧相位后验映射到内容空间作算术平均，独立于option排序控制；n442与32帧准确率差0.45pp不显著，origin flip率下降但order flip未解决；串行3.18s高于24帧2.86s，不能用attention单位冒充实测提速。", "review_level": "fulltext", "abstract_review_completed": true, "fulltext_review_completed": true, "evidence": [ { "url": "https://arxiv.org/abs/2610.08649", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08649v1", "location": "§3.2式2；§4 Table 1；§4.3" }, { "url": "https://arxiv.org/html/2610.08649v1", "location": "§3.2式2；§4 Table 1；§4.3", "notes": "PhaseFusion把3×8帧相位后验映射到内容空间作算术平均，独立于option排序控制；n442与32帧准确率差0.45pp不显著，origin flip率下降但order flip未解决；串行3.18s高于24帧2.86s，不能用attention单位冒充实测提速。" } ], "review_status": "reviewed", "official_url": "https://arxiv.org/html/2610.08649v1", "locations": [ "§3.2式2；§4 Table 1；§4.3" ], "implementation_status": "not_implemented", "fulltext_notes": "PhaseFusion把3×8帧相位后验映射到内容空间作算术平均，独立于option排序控制；n442与32帧准确率差0.45pp不显著，origin flip率下降但order flip未解决；串行3.18s高于24帧2.86s，不能用attention单位冒充实测提速。" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08647", "title": "SquidAgent: Parallelize Wisely, Coordinate Efficiently", "published": "2026-10-06T16:35:28Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08647", "title": "SquidAgent: Parallelize Wisely, Coordinate Efficiently", "published": "2026-10-06T16:35:28Z", "source_group": "original", "review_level": "fulltext", "source": "https://arxiv.org/html/2610.08647v1", "location": "§3.1–3.3 Eq1–10/Algorithm1；§4.1–4.6；Tables1–4正文讨论", "decision": "accepted", "reason": "SquidAgent一次DAG规划同时估算输出token/层对齐成本，session fork继承上下文、预写约定，以sum(tokens)/(max(tokens)+alignment)>alpha决定并行；明确可执行Agent调度机制。", "reviewer": "agent_review", "evidence": "同Sonnet4.6九任务吞吐17.2→38.1词/秒，质量rubric 97.1→98.2%；4任务去调度吞吐42.93→29.14；59子任务token预测Spearman0.77、时间预测0.16。", "limitations": "主要自建九任务、LLM rubric；token代理适用于生成主导场景不捕获tool/API时延；alpha1.9基于MathRef sweep设置，独立validation选择未建立；本地复现需真实LLM调用/成本不能fixture代替。", "fulltext_review_completed": true, "review_status": "reviewed", "implementation_status": "not_implemented_in_this_review" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08639", "title": "Forensic Reserve: Eliciting Latent Knowledge for Image Forgery Detection", "published": "2026-10-06T16:33:00Z", "reviews": [ { "track": "foundation-model", "id": "2610.08639", "title": "Forensic Reserve: Eliciting Latent Knowledge for Image Forgery Detection", "published": "2026-10-06T16:33:00Z", "repository_status": "new", "pool": "main", "decision": "accepted", "reason": "RGE用冻结DINOv3的site-wise ICA及real/fake discovery定位reserve、固定输出basis U，只训练零初始化A的残差UA h；classifier/backbone冻结，benchmark不参与discovery。可复用视觉adapter方法，当前证据仍image forensic专门任务。", "review_level": "fulltext", "abstract_review_completed": true, "fulltext_review_completed": true, "evidence": [ { "url": "https://arxiv.org/abs/2610.08639", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08639v1", "location": "§3.1；§3.2式3；§4.1" }, { "url": "https://arxiv.org/html/2610.08639v1", "location": "§3.1；§3.2式3；§4.1", "notes": "RGE用冻结DINOv3的site-wise ICA及real/fake discovery定位reserve、固定输出basis U，只训练零初始化A的残差UA h；classifier/backbone冻结，benchmark不参与discovery。可复用视觉adapter方法，当前证据仍image forensic专门任务。" } ], "review_status": "reviewed", "official_url": "https://arxiv.org/html/2610.08639v1", "locations": [ "§3.1；§3.2式3；§4.1" ], "implementation_status": "not_implemented", "fulltext_notes": "RGE用冻结DINOv3的site-wise ICA及real/fake discovery定位reserve、固定输出basis U，只训练零初始化A的残差UA h；classifier/backbone冻结，benchmark不参与discovery。可复用视觉adapter方法，当前证据仍image forensic专门任务。" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08630", "title": "Towards In-Parameter Memory Augmentation for Large Language Models", "published": "2026-10-06T16:27:47Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08630", "title": "Towards In-Parameter Memory Augmentation for Large Language Models", "published": "2026-10-06T16:27:47Z", "source_group": "original", "review_level": "abstract", "source": "https://arxiv.org/abs/2610.08630", "location": "Title and complete abstract", "decision": "out_of_scope", "reason": "参数记忆综述，无新核心算法。", "implementation_status": "not_implemented_in_this_review", "review_status": "reviewed" } ], "status": "rejected_or_out_of_scope", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08622", "title": "Agentic RCA for Internet-Scale Services Using Constrained Creativity", "published": "2026-10-06T16:21:57Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08622", "title": "Agentic RCA for Internet-Scale Services Using Constrained Creativity", "published": "2026-10-06T16:21:57Z", "source_group": "original", "review_level": "fulltext", "source": "https://arxiv.org/html/2610.08622v1", "location": "§3；§4.1–4.3；§6.1–6.5/Table3/Figs6–9；§8", "decision": "accepted", "reason": "E4以固定typed DAG语法约束RCA Agent，依次选playbook、组合已有operators、合成新operators；实际执行与admission形成可复现核心控制机制。", "reviewer": "agent_review", "evidence": "同Gemini3.1Pro对比LLM-only/RCA-Agent/Terminus2；DSB full0.91、starter-only0.72、无admission0.66；6生产incident pilot命中5例，其中2例需analyst hints。", "limitations": "§4.3用用户true-root-cause反馈admit，只能在训练/validation接收，不能test gold晋级；Accuracy@k,i允许多次尝试，不是单次accuracy；自建DSB有operator设计混杂；正文70 incident阶段统计与Table3总量不一致，图注≤56%与正文其他集0.80也不一致，headline不能无条件复用。", "institution": "Georgia Tech; Carnegie Mellon University; Conviva", "fulltext_review_completed": true, "review_status": "reviewed", "implementation_status": "not_implemented_in_this_review" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08621", "title": "Recursive Game Creator: An Agentic Product-Level Experience-Oriented Game Harness", "published": "2026-10-06T16:21:44Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08621", "title": "Recursive Game Creator: An Agentic Product-Level Experience-Oriented Game Harness", "published": "2026-10-06T16:21:44Z", "source_group": "original", "review_level": "fulltext", "source": "https://arxiv.org/html/2610.08621v1", "location": "§3.1–3.3；§4.1 Tables1–2；§4.2–4.4/Table3；§5", "decision": "accepted", "reason": "Recursive Game Creator包含designer/build、可执行多策略player rollout、盲化版本review及实际artifact retention，保留偏好证据到下一轮，属于固定模型Agent迭代闭环。", "reviewer": "agent_review", "evidence": "45 GameCraft任务同GPT6Astra baseline71.26→round3 77.89；47 GameASG任务success9/47→25/47，L2mean86.6→93.4。", "limitations": "非等预算基线，三轮与单轮混杂；用户study仅报告expert数值和playtime，未给强随机对照；97.1%coverage不是GUI速度提升；proxy player不代表人群偏好；能力复现必须实际运行游戏与policy不可gold固定轨迹。", "fulltext_review_completed": true, "review_status": "reviewed", "implementation_status": "not_implemented_in_this_review" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08586", "title": "MINDSET: Energy-based Schema Evolution for Long Conversational Agent Memory", "published": "2026-10-06T15:55:42Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08586", "title": "MINDSET: Energy-based Schema Evolution for Long Conversational Agent Memory", "published": "2026-10-06T15:55:42Z", "source_group": "original", "review_level": "fulltext", "source": "https://arxiv.org/html/2610.08586v1", "location": "§3.1–3.7；§4.1–4.4；§5.1–5.5；Limitations", "decision": "accepted", "reason": "MINDSET以immutable episodes和versioned schemas，按distortion/contradiction/history/fragmentation/inconsistency加权能量选择reinforce/supersede/split/new，再展开原始证据检索；定义完整可复现记忆算法。", "reviewer": "agent_review", "evidence": "LoCoMo700按conversation10簇显著提升排名；较LightMem F1+0.0256不显著p=.24023；Gemma上LightMem微领先0.0002；MAB150绝对ICL仅.024。", "limitations": "参数为开发集手调非最优理论，dev仅question-disjoint未证conversation-disjoint；LongContext按20k字节截断；基线为复刻，部分upstream artifacts缺失；最终stay margin而非switch margin决定行动，复现需保留这一实现细节。", "fulltext_review_completed": true, "review_status": "reviewed", "implementation_status": "not_implemented_in_this_review" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08560", "title": "Have I Seen Enough? Frozen Video-Language Models Encode Evidence Readiness", "published": "2026-10-06T15:43:04Z", "reviews": [ { "track": "foundation-model", "id": "2610.08560", "title": "Have I Seen Enough? Frozen Video-Language Models Encode Evidence Readiness", "published": "2026-10-06T15:43:04Z", "repository_status": "new", "pool": "main", "decision": "accepted", "reason": "Readiness Gating以冻结Qwen残差的嵌套CV logistic probe预测证据就绪，按训练分位阈值joint readiness/confidence首次停止或deadline最高readiness选答；视频级留出、crossed-quad消除图像/问题独立捷径。固定窗口policy不读证据时间，growing-prefix诊断采样依赖标注不能当部署延迟；§4.4给matched-video-seconds增益。", "review_level": "fulltext", "abstract_review_completed": true, "fulltext_review_completed": true, "evidence": [ { "url": "https://arxiv.org/abs/2610.08560", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08560v1", "location": "§3.1–3.3；§4.4；Appendix C/H/I" }, { "url": "https://arxiv.org/html/2610.08560v1", "location": "§3.1–3.3；§4.4；Appendix C/H/I", "notes": "Readiness Gating以冻结Qwen残差的嵌套CV logistic probe预测证据就绪，按训练分位阈值joint readiness/confidence首次停止或deadline最高readiness选答；视频级留出、crossed-quad消除图像/问题独立捷径。固定窗口policy不读证据时间，growing-prefix诊断采样依赖标注不能当部署延迟；§4.4给matched-video-seconds增益。" } ], "review_status": "reviewed", "official_url": "https://arxiv.org/html/2610.08560v1", "locations": [ "§3.1–3.3；§4.4；Appendix C/H/I" ], "implementation_status": "not_implemented", "fulltext_notes": "Readiness Gating以冻结Qwen残差的嵌套CV logistic probe预测证据就绪，按训练分位阈值joint readiness/confidence首次停止或deadline最高readiness选答；视频级留出、crossed-quad消除图像/问题独立捷径。固定窗口policy不读证据时间，growing-prefix诊断采样依赖标注不能当部署延迟；§4.4给matched-video-seconds增益。" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08539", "title": "RSJEV: Discriminative Remote Sensing Scene Classification with Multimodal Large Language Models", "published": "2026-10-06T15:28:36Z", "reviews": [ { "track": "foundation-model", "id": "2610.08539", "title": "RSJEV: Discriminative Remote Sensing Scene Classification with Multimodal Large Language Models", "published": "2026-10-06T15:28:36Z", "repository_status": "new", "pool": "main", "decision": "rejected", "reason": "RSJEV把已有JEV/Mapika式候选条件化单token判别读出及CE+Brier目标用于遥感分类；正文核查未发现区别于已登记LLM2Jev/既有Decider的新核心训练机制，作为专门领域复用不另立基础模型算法。", "review_level": "fulltext", "abstract_review_completed": true, "fulltext_review_completed": true, "evidence": [ { "url": "https://arxiv.org/abs/2610.08539", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08539v1", "location": "§I–III; §IV.A–E; TableII" }, { "url": "https://arxiv.org/html/2610.08539v1", "location": "§I–III; §IV.A–E; TableII", "notes": "以候选选项token的LM-head行直接读出概率；UCM/AID/NWPU三个已有数据集。Qwen3.5-0.8B 10epoch8A40、batch128，单A40延迟102.18ms，未提供新通用公共评测设施。" } ], "official_url": "https://arxiv.org/html/2610.08539v1", "locations": [ "§I–III", "§IV.A–E", "TableII" ], "code_url": "https://github.com/Dongtcs/RSJEV", "implementation_status": "not_implemented", "fulltext_notes": "以候选选项token的LM-head行直接读出概率；UCM/AID/NWPU三个已有数据集。Qwen3.5-0.8B 10epoch8A40、batch128，单A40延迟102.18ms，未提供新通用公共评测设施。", "review_status": "reviewed" } ], "status": "rejected_or_out_of_scope", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08528", "title": "MedCORE: Criteria-Grounded Clinical Reasoning for Interpretable Medical Image Diagnosis", "published": "2026-10-06T15:23:01Z", "reviews": [ { "track": "foundation-model", "id": "2610.08528", "title": "MedCORE: Criteria-Grounded Clinical Reasoning for Interpretable Medical Image Diagnosis", "published": "2026-10-06T15:23:01Z", "repository_status": "new", "pool": "main", "decision": "out_of_scope", "reason": "MedCORE面向皮肤、超声、视网膜病理诊断的临床标准图网络；属于特定医疗应用而非通用基础模型算法。", "review_level": "abstract", "abstract_review_completed": true, "fulltext_review_completed": false, "evidence": [ { "url": "https://arxiv.org/abs/2610.08528", "location": "Abstract（完整摘要已逐条审查）" } ] } ], "status": "rejected_or_out_of_scope", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08526", "title": "WareFly-VLA: A Vision-Language-Action Framework for UAV Navigation and Human Tracking in Smart Warehouses", "published": "2026-10-06T15:21:14Z", "reviews": [ { "track": "foundation-model", "id": "2610.08526", "title": "WareFly-VLA: A Vision-Language-Action Framework for UAV Navigation and Human Tracking in Smart Warehouses", "published": "2026-10-06T15:21:14Z", "repository_status": "new", "pool": "main", "decision": "pending", "reason": "WareFly-VLA定义507飞行轨迹、4DoF连续动作及431/76 episode级固定split，有独立UAV评测价值；§6仅ground-truth帧上的open-loop MAE/correlation及轨迹积分，非闭环飞行成功率。§Data availability明确数据/runner仍access on request、acceptance后公开，故等待公开入口而非未读完。", "review_level": "fulltext", "kind": "evaluation_scope_recheck", "abstract_review_completed": true, "fulltext_review_completed": true, "evidence": [ { "url": "https://arxiv.org/abs/2610.08526", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08526v1", "location": "§3；§5.2；§6.1 Tables 5–6；§7 limitations；Data availability" }, { "url": "https://arxiv.org/html/2610.08526v1", "location": "§3；§5.2；§6.1 Tables 5–6；§7 limitations；Data availability", "notes": "WareFly-VLA定义507飞行轨迹、4DoF连续动作及431/76 episode级固定split，有独立UAV评测价值；§6仅ground-truth帧上的open-loop MAE/correlation及轨迹积分，非闭环飞行成功率。§Data availability明确数据/runner仍access on request、acceptance后公开，故等待公开入口而非未读完。" } ], "review_status": "reviewed_deferred", "official_url": "https://arxiv.org/html/2610.08526v1", "locations": [ "§3；§5.2；§6.1 Tables 5–6；§7 limitations；Data availability" ], "implementation_status": "not_implemented", "fulltext_notes": "WareFly-VLA定义507飞行轨迹、4DoF连续动作及431/76 episode级固定split，有独立UAV评测价值；§6仅ground-truth帧上的open-loop MAE/correlation及轨迹积分，非闭环飞行成功率。§Data availability明确数据/runner仍access on request、acceptance后公开，故等待公开入口而非未读完。" } ], "status": "reviewed_deferred", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08514", "title": "How Much Evidence Should a Coding Agent's Self-Correction Carry? Adaptive Dirichlet Evidence for Self-Distillation", "published": "2026-10-06T15:17:56Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08514", "title": "How Much Evidence Should a Coding Agent's Self-Correction Carry? Adaptive Dirichlet Evidence for Self-Distillation", "published": "2026-10-06T15:17:56Z", "source_group": "original", "review_level": "fulltext", "source": "https://arxiv.org/html/2610.08514v1", "location": "§3.1–3.3 Eq1–11；§4.1–4.4；§5", "decision": "accepted", "reason": "EESD将公开执行before/after四种转移和词法相关性转为有效样本量1/sum(p²)，Dirichlet utility lower score加权CE并保留未加权forward KL；可复现Agent自修正训练机制。", "reviewer": "agent_review", "evidence": "DeepSeek/CodeARC fresh Pass@1 15.0→20.4%，配对source CI[2.8,8.0]pp；12设置范围−6.8到+5.4pp。", "limitations": "只一轮同source适应，不是source-disjoint泛化；full update vs no-update不能归因有效质量或KL；bootstrap条件于单次训练；有效质量仅relevance集中度，不自动修复执行依赖。", "fulltext_review_completed": true, "review_status": "reviewed", "implementation_status": "not_implemented_in_this_review" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08513", "title": "Wiki-Talkie: Multilingual Benchmarking of Persona-Based Agents on Real-World Discussions", "published": "2026-10-06T15:17:46Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08513", "title": "Wiki-Talkie: Multilingual Benchmarking of Persona-Based Agents on Real-World Discussions", "published": "2026-10-06T15:17:46Z", "source_group": "original", "review_level": "fulltext", "source": "https://arxiv.org/html/2610.08513v1", "location": "§3.1；§4；§5.2；Limitations/Ethics；official GitHub API2026-10-07", "decision": "pending", "reason": "Amazon Wiki-Talkie的五语人口行为分布评测有公共评测价值，但官方amazon-science/wikitalkie返回404，数据又明确申请访问非开放镜像；模型A–D及extractor匿名，现阶段审结待资源。", "implementation_status": "not_implemented_in_this_review", "review_status": "reviewed_deferred", "reviewer": "agent_review", "evidence": "Wiki2025-10dumps，文章取首评论前revision；next-turn使用真实历史，history5–20先前评论，dialogact annotatorF1.76，JSD/情绪WD衡量群体而非个体人格；CCBYSA4数据声明。", "limitations": "behavior从用户50随机threads提炼，接入需重新核严格时间/目标排除；history与profile非信息量配平，profile长度分层不能证明因果；模型匿名阻原数值复现；saltedhash不消除文本反识别，群体研究不得个体冒充。", "code": "https://github.com/amazon-science/wikitalkie", "pending_reason": "official_repository_404_access_request_data_anonymous_models", "candidate_kind": "evaluation_infrastructure", "fulltext_review_completed": true } ], "status": "reviewed_deferred", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08482", "title": "Knee3DVLM: Dual-Sequence Full-Volume Vision-Language Modeling for Comprehensive Knee MRI Assessment", "published": "2026-10-06T14:58:53Z", "reviews": [ { "track": "foundation-model", "id": "2610.08482", "title": "Knee3DVLM: Dual-Sequence Full-Volume Vision-Language Modeling for Comprehensive Knee MRI Assessment", "published": "2026-10-06T14:58:53Z", "repository_status": "new", "pool": "main", "decision": "out_of_scope", "reason": "Knee3DVLM为膝关节MRI双序列57诊断目标建模，属特定医疗模型应用。", "review_level": "abstract", "abstract_review_completed": true, "fulltext_review_completed": false, "evidence": [ { "url": "https://arxiv.org/abs/2610.08482", "location": "Abstract（完整摘要已逐条审查）" } ] } ], "status": "rejected_or_out_of_scope", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08463", "title": "UNREAL: Unifying Retrieval and Long-Context with a Single Model", "published": "2026-10-06T14:48:53Z", "reviews": [ { "track": "foundation-model", "id": "2610.08463", "title": "UNREAL: Unifying Retrieval and Long-Context with a Single Model", "published": "2026-10-06T14:48:53Z", "repository_status": "new", "pool": "main", "decision": "accepted", "reason": "UNREAL用冻结decoder中层chunk表征、训练64检索token及跨层混合做MaxSim，再重编码所选文本生成；提供multi-positive InfoNCE、公开QA训练split与long-context独立验证协议。", "review_level": "fulltext", "abstract_review_completed": true, "fulltext_review_completed": true, "evidence": [ { "url": "https://arxiv.org/abs/2610.08463", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08463v1", "location": "§2.1–2.3 Eq1–7; §3/Table1; §4.2/Figure7; §6; Appendix B.1–B.3" }, { "url": "https://arxiv.org/html/2610.08463v1", "location": "§2.1–2.3 Eq1–7; §3/Table1; §4.2/Figure7; §6; Appendix B.1–B.3", "notes": "21Mchunk 3BtokenWiki2018，query初始top5 BM25，每chunk7vectors、query4groups、<500K训练参数；15K步batch256，500hardnegative从训练gold挖掘。H100约32K后测得TTFT优势。" } ], "official_url": "https://arxiv.org/html/2610.08463v1", "locations": [ "§2.1–2.3 Eq1–7", "§3/Table1", "§4.2/Figure7", "§6", "Appendix B.1–B.3" ], "limitations": "测试不可注入oracle，只有训练hardnegative挖掘用gold；100M场景以稀疏证据为主，dashed延迟是FLOP外推非实测。索引离线编码存储和两次模型forward须计成本；不同baseline训练预算并非完全匹配。", "implementation_status": "not_implemented", "fulltext_notes": "21Mchunk 3BtokenWiki2018，query初始top5 BM25，每chunk7vectors、query4groups、<500K训练参数；15K步batch256，500hardnegative从训练gold挖掘。H100约32K后测得TTFT优势。", "review_status": "reviewed" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08452", "title": "Agentic AutoRAG: RAG Pipeline Optimization through Reasoning-Driven Agents", "published": "2026-10-06T14:38:15Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08452", "title": "Agentic AutoRAG: RAG Pipeline Optimization through Reasoning-Driven Agents", "published": "2026-10-06T14:38:15Z", "source_group": "original", "review_level": "fulltext", "source": "https://arxiv.org/html/2610.08452v1", "location": "§3.1–3.3；§4；§5/Table2；Limitations", "decision": "accepted", "reason": "Agentic AutoRAG的检索/生成归因→Diagnoser→KB/成本引导Proposer构成明确搜索机制，可接入evolve候选；准确率实验有100验证/300独立heldout隔离。", "implementation_status": "not_implemented_in_this_review", "review_status": "reviewed", "reviewer": "agent_review", "evidence": "冻结exam+gold-span命中仅用于验证优化；30trial×10seed，heldout judge Hotpot/MuSiQue/MultiHop .825/.330/.790；生成exam的Pareto医疗场景77%vs71.5%，成本57.7%。", "limitations": "医疗Pareto只报告反复优化的同一exam，非heldout能力；成本轴只含生成/扩展API，不含本地embedding/rerank；noKB/diag合并消融不能归因单组件；warmTPE获免费30次同库先验，且不是完整syftr；judge无人类一致性验证。", "code": "https://github.com/Agentic-Systems-Lab/Agentic-AutoRAG", "code_status": "official GitHub contents lists source/configs/knowledge_base/scripts/tests/LICENSE; not executed", "fulltext_review_completed": true } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08430", "title": "NeMo-DCR: Bit-Exact Delta-Compressed Refit for Scalable Agentic RL at Trillion-Parameter Scale", "published": "2026-10-06T14:30:06Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08430", "title": "NeMo-DCR: Bit-Exact Delta-Compressed Refit for Scalable Agentic RL at Trillion-Parameter Scale", "published": "2026-10-06T14:30:06Z", "source_group": "original", "review_level": "abstract", "source": "https://arxiv.org/abs/2610.08430", "location": "Title and complete abstract", "decision": "out_of_scope", "reason": "NeMo-DCR权重同步/精确delta传输系统，归训练基础设施而非Agent策略。", "implementation_status": "not_implemented_in_this_review", "review_status": "reviewed" } ], "status": "rejected_or_out_of_scope", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08414", "title": "Image Bitstream Fine-grained Understanding for Privacy-Friendly AIoT", "published": "2026-10-06T14:18:54Z", "reviews": [ { "track": "foundation-model", "id": "2610.08414", "title": "Image Bitstream Fine-grained Understanding for Privacy-Friendly AIoT", "published": "2026-10-06T14:18:54Z", "repository_status": "new", "pool": "main", "decision": "accepted", "reason": "BFG将JPEG bytes经ByteFormer分层卷积/shifted-window编码为192维，projector到512维后全局上下文融合+局部cross-attention自回归描述，另120类head；CFU-D源ID/split隔离并测试替换/删除/重复腐坏。Dogs分类58.51%采用ImageNet初始化；不解码不等价正式隐私保证，不能包装为通用大规模foundation结论。", "review_level": "fulltext", "abstract_review_completed": true, "fulltext_review_completed": true, "evidence": [ { "url": "https://arxiv.org/abs/2610.08414", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08414v1", "location": "§III-B式2–5；§III-C；§IV-A–D Tables I/IV/V" }, { "url": "https://arxiv.org/html/2610.08414v1", "location": "§III-B式2–5；§III-C；§IV-A–D Tables I/IV/V", "notes": "BFG将JPEG bytes经ByteFormer分层卷积/shifted-window编码为192维，projector到512维后全局上下文融合+局部cross-attention自回归描述，另120类head；CFU-D源ID/split隔离并测试替换/删除/重复腐坏。Dogs分类58.51%采用ImageNet初始化；不解码不等价正式隐私保证，不能包装为通用大规模foundation结论。" } ], "review_status": "reviewed", "official_url": "https://arxiv.org/html/2610.08414v1", "locations": [ "§III-B式2–5；§III-C；§IV-A–D Tables I/IV/V" ], "implementation_status": "not_implemented", "fulltext_notes": "BFG将JPEG bytes经ByteFormer分层卷积/shifted-window编码为192维，projector到512维后全局上下文融合+局部cross-attention自回归描述，另120类head；CFU-D源ID/split隔离并测试替换/删除/重复腐坏。Dogs分类58.51%采用ImageNet初始化；不解码不等价正式隐私保证，不能包装为通用大规模foundation结论。" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08407", "title": "Seeing the Context: Enhancing Recommender Systems with Image-Derived Contextual Signals", "published": "2026-10-06T14:15:48Z", "reviews": [ { "track": "recommendation", "id": "2610.08407", "title": "Seeing the Context: Enhancing Recommender Systems with Image-Derived Contextual Signals", "published": "2026-10-06T14:15:48Z", "repository_status": "new", "decision": "rejected", "review_level": "full-text", "fulltext_review_completed": true, "reason": "ICE-Fuse 在 TripAdvisor 图文上下文上做 RMSE/消融；未提供线上分流或部署实验，不满足工业搜广推准入。", "evidence": [ { "url": "https://arxiv.org/abs/2610.08407", "location": "§4.1–4.3 Tables 2–3; §5" } ], "fulltext_receipt": { "id": "2610.08407", "url": "https://arxiv.org/html/2610.08407v1", "sha256": "2223d22274d906bfa43f36d184feed98ec023a0af2ebe0b83ce2c0f97262ed78", "status": "downloaded", "characters": 28755, "errors": [] } } ], "status": "rejected_or_out_of_scope", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08403", "title": "SSR: Sparse Segment Reduction for Ternary GEMM Acceleration", "published": "2026-10-06T14:15:12Z", "reviews": [ { "track": "foundation-model", "id": "2610.08403", "title": "SSR: Sparse Segment Reduction for Ternary GEMM Acceleration", "published": "2026-10-06T14:15:12Z", "repository_status": "new", "pool": "main", "decision": "accepted", "reason": "SSR 将三值矩阵分成正负二值矩阵，按列块模式排序并剔除全零模式，以分段求和与树规约代替乘法；公开 C++/AVX2 实现可核验。", "review_level": "fulltext", "abstract_review_completed": true, "fulltext_review_completed": true, "evidence": [ { "url": "https://arxiv.org/abs/2610.08403", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08403v1", "location": "§III-A/B Algorithms1–2; §IV-A/B; §IV-C TableIII; §V" }, { "url": "https://arxiv.org/html/2610.08403v1", "location": "§III-A/B Algorithms1–2; §IV-A/B; §IV-C TableIII; §V", "notes": "" } ], "kind": "cpu_inference_kernel", "locations": [ "§III-A/B Algorithms1–2", "§IV-A/B", "§IV-C TableIII", "§V" ], "limitations": "报告的是 CPU，部分高稀疏设置慢于 CSC/Eigen；原文未验证 GPU，不能用仓库其他项目的 GPU 目录冒充 SSR。模型级主要是改造三值 MLP 的延迟，没有公开质量保真证据。作者 README 引用 DATE2026，应标记本次 arXiv 收录而非确认首次发表。", "code_url": "https://github.com/fpgasystems/ternaryLLM/tree/main/SSR", "code_status": "author-linked repository accessible; MIT; SSR is separate from ternaryLLM_GPU", "official_url": "https://arxiv.org/html/2610.08403v1", "review_provenance": "root 独立正文审查；foundation读取整合", "implementation_status": "not_implemented", "fulltext_notes": "", "review_status": "reviewed" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08401", "title": "GeoPID: Decomposing and Steering Visual Information in Vision-Language Models", "published": "2026-10-06T14:14:25Z", "reviews": [ { "track": "foundation-model", "id": "2610.08401", "title": "GeoPID: Decomposing and Steering Visual Information in Vision-Language Models", "published": "2026-10-06T14:14:25Z", "repository_status": "new", "pool": "main", "decision": "accepted", "reason": "GeoPID从无标签校准表征估计模态共享/独占子空间，推理按I+αUUᵀ放大vision-unique；22×14组合每组100道独立校准题用标签选层/α后冻结，最终评测不读gold。任务Rényi信息分解为离线诊断，平均相对7.63%需保留308组异质/部分负收益。", "review_level": "fulltext", "abstract_review_completed": true, "fulltext_review_completed": true, "evidence": [ { "url": "https://arxiv.org/abs/2610.08401", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08401v1", "location": "§2.1–2.3式11；§3.1；§3.2 Table 5" }, { "url": "https://arxiv.org/html/2610.08401v1", "location": "§2.1–2.3式11；§3.1；§3.2 Table 5", "notes": "GeoPID从无标签校准表征估计模态共享/独占子空间，推理按I+αUUᵀ放大vision-unique；22×14组合每组100道独立校准题用标签选层/α后冻结，最终评测不读gold。任务Rényi信息分解为离线诊断，平均相对7.63%需保留308组异质/部分负收益。" } ], "review_status": "reviewed", "official_url": "https://arxiv.org/html/2610.08401v1", "locations": [ "§2.1–2.3式11；§3.1；§3.2 Table 5" ], "implementation_status": "not_implemented", "fulltext_notes": "GeoPID从无标签校准表征估计模态共享/独占子空间，推理按I+αUUᵀ放大vision-unique；22×14组合每组100道独立校准题用标签选层/α后冻结，最终评测不读gold。任务Rényi信息分解为离线诊断，平均相对7.63%需保留308组异质/部分负收益。" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08400", "title": "Atom-JEPA: Joint-Embedding Predictive Architecture for 3D Atomistic Systems", "published": "2026-10-06T14:13:51Z", "reviews": [ { "track": "foundation-model", "id": "2610.08400", "title": "Atom-JEPA: Joint-Embedding Predictive Architecture for 3D Atomistic Systems", "published": "2026-10-06T14:13:51Z", "repository_status": "new", "pool": "main", "decision": "pending", "reason": "Atom-JEPA具备真实SE(3)等变原子/子结构预测算法与公开迁移结果，但严格等预算新目标归因尚未满足：§G.1同Uni-Mol atom-only与atom+pool的批量384/128、学习率2.42e-4/1.4e-4不同；§B未做同架构其他SSL比较。需明确可接受的匹配控制或补齐预算证据，不能把scratch对比当等预训练预算。", "review_level": "fulltext", "official_url": "https://arxiv.org/html/2610.08400v1", "locations": [ "§3", "§4.2/Table4", "Appendix B", "Appendix G.1/Table26", "Appendix H.1.1" ], "evidence": [ { "url": "https://arxiv.org/abs/2610.08400", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08400v1", "location": "§3; §4.2/Table4; Appendix B; Appendix G.1/Table26; Appendix H.1.1" }, { "url": "https://arxiv.org/html/2610.08400v1", "location": "§3; §4.2/Table4; Appendix B; Appendix G.1/Table26; Appendix H.1.1", "notes": "19M Uni-Mol与1.7M Alexandria预训练，双目标EMA EquiformerV3；Matbench40任务fold全部优于scratch，ChEMBL-MT仍有-0.5%负结果。Uni-Mol2475 GPU小时、Alexandria2147 GPU小时；主目标对照训练配置不相同。" } ], "code_url": "https://github.com/khelverskovp/atom-jepa", "institution": "Technical University of Denmark", "implementation_status": "not_implemented", "abstract_review_completed": true, "fulltext_review_completed": true, "fulltext_notes": "19M Uni-Mol与1.7M Alexandria预训练，双目标EMA EquiformerV3；Matbench40任务fold全部优于scratch，ChEMBL-MT仍有-0.5%负结果。Uni-Mol2475 GPU小时、Alexandria2147 GPU小时；主目标对照训练配置不相同。", "review_status": "reviewed_deferred" } ], "status": "reviewed_deferred", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08364", "title": "Transect: Retaining Observability for Long-Horizon LLM Agent Evaluations", "published": "2026-10-06T13:52:15Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08364", "title": "Transect: Retaining Observability for Long-Horizon LLM Agent Evaluations", "published": "2026-10-06T13:52:15Z", "source_group": "original", "review_level": "fulltext", "source": "https://arxiv.org/html/2610.08364v1", "location": "§2.1–2.2；§3.1/3.3–3.5；§4；official README/API", "decision": "accepted", "reason": "Transect为公共可执行评测观测工具：family Spec固定词表，structural/judged scanners→provenance tables→源链接报告；可用于evolve审计但不是任务能力评分器。", "implementation_status": "not_implemented_in_this_review", "review_status": "reviewed", "reviewer": "agent_review", "evidence": "UK AISI/Meridian MIT公开repo，Python3.12+，structural worked example无judge调用；solo/kroll/cohort记录member votes/verifier修改。单CRUX案例580orchestrator outputs/131sessions922fileaccess，来源/覆盖明确。", "limitations": "仅单案例描述，agreement/repeatability不是正确率；读文件不证明采纳、sourceorder非walltime、compaction邻近非因果；Inspect machine-global缓存可能伪造重复稳定性，独立judge重复须freshcache；扩展扫描器可能自行调用模型。", "code": "https://github.com/AI-Safety-Institute/transect", "code_status": "MIT; executable structural example and judged entrypoints verified in README, not executed", "candidate_kind": "evaluation_observability_infrastructure", "fulltext_review_completed": true } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08341", "title": "DIPrune: Task-Aware Token Pruning with Dual Importance for Efficient Multimodal Language Models", "published": "2026-10-06T13:38:27Z", "reviews": [ { "track": "foundation-model", "id": "2610.08341", "title": "DIPrune: Task-Aware Token Pruning with Dual Importance for Efficient Multimodal Language Models", "published": "2026-10-06T13:38:27Z", "repository_status": "new", "pool": "main", "decision": "accepted", "reason": "DIPrune结合文本到视觉注意力倒置rank归一与跨层rank上升的截断归一，保留早层低但后层新出现语义token并移除对应KV；多VLM及视频骨干评测、层对齐消融。LLaVA64tokens平均保留97.7%，POPE TTFT73.23→32.50ms、额外0.24ms，需以实际TTFT而非FLOPs主张速度。", "review_level": "fulltext", "abstract_review_completed": true, "fulltext_review_completed": true, "evidence": [ { "url": "https://arxiv.org/abs/2610.08341", "location": "完整摘要逐条审查" }, { "url": "https://arxiv.org/html/2610.08341v1", "location": "§5.1式6–7；§5.2式8；§6.1；§6.5 Table 4" }, { "url": "https://arxiv.org/html/2610.08341v1", "location": "§5.1式6–7；§5.2式8；§6.1；§6.5 Table 4", "notes": "DIPrune结合文本到视觉注意力倒置rank归一与跨层rank上升的截断归一，保留早层低但后层新出现语义token并移除对应KV；多VLM及视频骨干评测、层对齐消融。LLaVA64tokens平均保留97.7%，POPE TTFT73.23→32.50ms、额外0.24ms，需以实际TTFT而非FLOPs主张速度。" } ], "review_status": "reviewed", "official_url": "https://arxiv.org/html/2610.08341v1", "locations": [ "§5.1式6–7；§5.2式8；§6.1；§6.5 Table 4" ], "implementation_status": "not_implemented", "fulltext_notes": "DIPrune结合文本到视觉注意力倒置rank归一与跨层rank上升的截断归一，保留早层低但后层新出现语义token并移除对应KV；多VLM及视频骨干评测、层对齐消融。LLaVA64tokens平均保留97.7%，POPE TTFT73.23→32.50ms、额外0.24ms，需以实际TTFT而非FLOPs主张速度。" } ], "status": "accepted", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08337", "title": "Machine Learning for German Redispatch Forecasting under Data Delays and Temporal Distribution Shift", "published": "2026-10-06T13:35:50Z", "reviews": [ { "track": "foundation-model", "id": "2610.08337", "title": "Machine Learning for German Redispatch Forecasting under Data Delays and Temporal Distribution Shift", "published": "2026-10-06T13:35:50Z", "repository_status": "new", "pool": "main", "decision": "out_of_scope", "reason": "德国电网再调度概率预测比较，属于能源时序应用而非基础模型核心算法。", "review_level": "abstract", "abstract_review_completed": true, "fulltext_review_completed": false, "evidence": [ { "url": "https://arxiv.org/abs/2610.08337", "location": "Abstract（完整摘要已逐条审查）" } ] } ], "status": "rejected_or_out_of_scope", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08331", "title": "Transferable Spatial Temporal Coherence Adversarial Attack on Black-Box Vision Language Models for Autonomous Driving", "published": "2026-10-06T13:30:06Z", "reviews": [ { "track": "foundation-model", "id": "2610.08331", "title": "Transferable Spatial Temporal Coherence Adversarial Attack on Black-Box Vision Language Models for Autonomous Driving", "published": "2026-10-06T13:30:06Z", "repository_status": "new", "pool": "main", "decision": "out_of_scope", "reason": "自动驾驶视频VLM迁移攻击研究，主贡献为攻击构造而非基础模型能力训练/推理方法。", "review_level": "abstract", "abstract_review_completed": true, "fulltext_review_completed": false, "evidence": [ { "url": "https://arxiv.org/abs/2610.08331", "location": "Abstract（完整摘要已逐条审查）" } ] } ], "status": "rejected_or_out_of_scope", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08327", "title": "MedZERO: Self-Evolving Agents for Open-Ended Medical Reasoning Through Controlled Knowledge Accumulation", "published": "2026-10-06T13:25:55Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08327", "title": "MedZERO: Self-Evolving Agents for Open-Ended Medical Reasoning Through Controlled Knowledge Accumulation", "published": "2026-10-06T13:25:55Z", "source_group": "original", "review_level": "fulltext", "source": "https://arxiv.org/html/2610.08327v1", "location": "§3 Eq1–21；§4.1–4.2；§5；AppendixB；Open access checklist", "decision": "pending", "reason": "MedZERO交替训练Examiner/Reasoner与临时→持久KG准入是合格Agent机制；正文明确投稿时不发布代码、计划出版后发布，grounding scorer/映射阈值等关键实现不足，记审结待发布。", "implementation_status": "not_implemented_in_this_review", "review_status": "reviewed_deferred", "reviewer": "agent_review", "evidence": "4响应option映射多数一致性作为弱frontier，curation保留[.3,.8]；临时triplet仅可读前批且最终丢弃，独立重生成去重持久化；8B六分项均值53.5vsAgent043.7。", "limitations": "多数一致性不是医学正确率；持久准入缺独立事实验证，仍可能自我强化错误；MCQ选项只给后置judge不应给被测Reasoner；临床效用未建立，KG/web/judge依赖明显，正文未充分规定grounding评分函数。", "pending_reason": "author_code_not_released_and_core_details_incomplete", "fulltext_review_completed": true } ], "status": "reviewed_deferred", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08300", "title": "Memory Depth and Reconstructed Context Width: A Controlled Evaluation of Hierarchical Retrieval", "published": "2026-10-06T13:04:43Z", "reviews": [ { "track": "foundation-model", "id": "2610.08300", "title": "Memory Depth and Reconstructed Context Width: A Controlled Evaluation of Hierarchical Retrieval", "published": "2026-10-06T13:04:43Z", "repository_status": "new", "pool": "main", "decision": "out_of_scope", "reason": "分层会话检索深度/上下文宽度的控制实验，主要为系统诊断而非新的基础模型核心算法。", "review_level": "abstract", "abstract_review_completed": true, "fulltext_review_completed": false, "evidence": [ { "url": "https://arxiv.org/abs/2610.08300", "location": "Abstract（完整摘要已逐条审查）" } ] } ], "status": "rejected_or_out_of_scope", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08250", "title": "MASC: A Multi-Agent Self-Calibration Framework with Latent Construct Alignment for Consistent Client Role-Playing in Psychological Counseling", "published": "2026-10-06T12:31:05Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08250", "title": "MASC: A Multi-Agent Self-Calibration Framework with Latent Construct Alignment for Consistent Client Role-Playing in Psychological Counseling", "published": "2026-10-06T12:31:05Z", "source_group": "original", "review_level": "fulltext", "source": "https://arxiv.org/html/2610.08250v1", "location": "Multi-Agent Self-Calibration §方法 Eq1–16/Alg1；Experiment Tables2–3；官方GitHub API", "decision": "pending", "reason": "MASC以预测情绪/状态/动作、三Agent五步辩论、投票一致性优先和两轮失败记忆重试维持角色；方法与CRPC评测有关联价值，但官方仓库description明确code and data coming soon且无license。", "implementation_status": "not_implemented_in_this_review", "review_status": "reviewed_deferred", "reviewer": "agent_review", "evidence": "38profiles；只读过去对话不读未来reference；最多3尝试、每次3×5generation；独立盲评而非复用内部投票标签；Hetero action/emotion/state KL .06/.27/.06。", "limitations": "38profiles及同套构念/提示限制外推；Hetero与Homo骨干不同不能隔离多样性原因；73.97%最终匹配，44.19%是以最终收敛为条件的重试占比非修复概率；KL仅群体标签频率不是治疗真实性。", "code": "https://github.com/DeSun99/MASC-CRPC", "pending_reason": "author_code_and_data_coming_soon", "fulltext_review_completed": true } ], "status": "reviewed_deferred", "implementation_status": "not_implemented_in_this_review", "inherited_review_artifact": "research-audits/oct07-review-closure/paper-register.json" }, { "id": "2610.08246", "title": "LeanPlan: Optimal Planning with LLM-Generated Heuristics and Admissibility Proofs", "published": "2026-10-06T12:27:34Z", "reviews": [ { "track": "agent", "arxiv_id": "2610.08246", "title": "LeanPlan: Optimal Planning with LLM-Generated Heuristics and Admissibility Proofs", "published": "2026-10-06T12:27:34Z", "source_group": "original", "review_level": "fulltext", "source": "https://arxiv.org/html/2610.08246v1", "location": "§3–5；§7 Limitations；Reproducibility", "decision": "pending", "reason": "LeanPlan生成Lean启发式+可采纳性证明，controller拒绝sorry/admit/native_decide/新公理且独立重编译与训练任务验收，符合可验证Agent/evolve机制；作者明确代码/benchmark/log接受后才发布，审结待发布。", "implementation_status": "not_implemented_in_this_review", "review_status": "reviewed_deferred", "reviewer": "agent_review", "evidence": "13domain1170test；sandbox隔离test，按domain单次搜索均69.8分钟17.30美元；595超训练对象规模test解53vsSCP13；每任务300CPU秒8GiB。", "limitations": "SCP预算曾用test suite调优，原对照非严格完整test隔离；每domain单次Agent生成，无多seed；证明仅对certificate通过且成本