{ "schema_version": 1, "track": "recommendation", "reviewed_on": "2026-10-15", "reviews": [ { "id": "2610.20578", "title": "Creating SquashFS Images to Support HPC and AI Datasets", "published": "2026-10-14T17:56:16Z", "revision": 1, "priority": "P2", "decision": "rejected_or_out_of_scope", "resolution": "outside_model_algorithm_scope", "reason": "比较十一种 SquashFS 数据镜像打包流程，贡献是 HPC 文件系统操作指南，不是检索推荐模型、LLM、后训练或 Agent 算法。", "review_level": "abstract_and_bounded_original_scope_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.20578v1", "source_sha256": "03204ca20337bcafe6247b7bbc9015b2d64585e67fe26c9a16e981aef2775e2a", "metadata_sha256": "361d6b6cc07354c09673fe499b8332dd861217f169ffa4d30ef0e4c2971307d3", "locations": [ "abstract", "bounded defining method / experimental setup / evaluation / limitations" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.20191", "title": "HEM-Med: A Hierarchical Agentic Memory Framework for Safe Multi-Center Medication Recommendation", "published": "2026-10-14T13:39:34Z", "revision": 1, "priority": "P2", "decision": "rejected_or_out_of_scope", "resolution": "outside_industrial_recommendation_scope", "reason": "HEM-Med 解耦跨医院经验与本地处方记忆，配安全集合解码；eICU 多中心离线医疗推荐不等于互联网搜广推业务 A/B，也不是通用 Agent 机制，本轮不纳入。", "review_level": "abstract_and_bounded_original_scope_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.20191v1", "source_sha256": "a7ad18d2ec0fd419986dcfe0351ddc65f101e3e6bf72ab01b7629e57a7df39e7", "metadata_sha256": "2909d1c50870b1d93df369f3d61264167fef021d2df2b89483f0cc62a5a7e49b", "locations": [ "abstract", "bounded defining method / experimental setup / evaluation / limitations" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.19578", "title": "Bridging the Topological Divide: Dual-Branch Graph Distillation with Semantic Priors for Robust Cold-Start Recommendation", "published": "2026-10-14T04:43:38Z", "revision": 1, "priority": "P2", "decision": "rejected_or_out_of_scope", "resolution": "industrial_evidence_gate_not_met", "reason": "DualGCN 两阶段图教师/语义学生、余弦蒸馏与范数重建解决冷启动；正文实验为三个公开数据集离线评测，没有该方法的生产业务 A/B，不增列经典例外。", "review_level": "abstract_and_bounded_original_scope_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.19578v1", "source_sha256": "7fa701705e9ba1ecc04b8f8b92318343c760347aefcb392d930fef1e9b3eaef9", "metadata_sha256": "ba12ed0b6e8ea6bf49ce81f162017efe035c2168a022884e281eb32572ca843d", "locations": [ "abstract", "bounded defining method / experimental setup / evaluation / limitations" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.19185", "title": "How to Evaluate When You Don't Know How to Evaluate? Choosing Reliable Setups for Base Language Model Benchmarking", "published": "2026-10-13T20:50:28Z", "revision": 1, "priority": "P1", "decision": "reviewed_deferred", "resolution": "not_reproduced_high_cost", "reason": "用单调性、非随机性和排序一致性选择 LLM 评测设置；Appendix A 实报 2B 模型约11000、较大模型约7000 GH200 GPU-hours。CPU 五分钟仅是已存在结果的元指标计算，不是整套评测成本；不复现、审查 DONE。", "review_level": "abstract_and_bounded_original_scope_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.19185v1", "source_sha256": "e209eaac6b47512baefad6a141d8fcbed92c5b145b8f434d024e363ddc1515ef", "metadata_sha256": "5e1dd18c0618725b8697388a096db116fa7151b8ffcc49996f764e6983cc7ea6", "locations": [ "Appendix A: Inference Details" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix.", "cost_assessment": { "source_url": "https://arxiv.org/html/2610.19185v1", "basis": "Appendix A: Inference Details", "scale": "约18000 GH200 GPU-hours；26个2B模型、80任务、4语言及较大模型", "available_resource": "单卡 A100 或 A30，或本地 Mac；不假设多卡同时可用", "cost_type": "reported" } }, { "id": "2610.19065", "title": "Revisiting Generative vs. Dense Retrieval: A Static Value Prior Reverses the Gap", "published": "2026-10-13T19:02:30Z", "revision": 1, "priority": "P2", "decision": "rejected_or_out_of_scope", "resolution": "industrial_evidence_gate_not_met", "reason": "比较 GR/DR 的 exact target、替代品和全列表价值，静态 item value 可改变比较顺序；§6.3 工业数据仍是 LLM-judged relevance 离线实验，并非真实用户生产 A/B。保留指标警示，不作工业实现候选。", "review_level": "abstract_and_bounded_original_scope_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.19065v1", "source_sha256": "d502b6473a740e5ad13be8eb60a9df037ec8e08374b3482fb76d6fd71db001bc", "metadata_sha256": "15a749762d10a1c5f206ee06f5ed1d5ded96ff92c1b5f8eea8011c7d16452ebd", "locations": [ "abstract", "bounded defining method / experimental setup / evaluation / limitations" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.18465", "title": "How Safe Is ChatGPT Health Advice? Evidence from Real User Conversations in Four Global South Countries", "published": "2026-10-13T14:30:12Z", "revision": 1, "priority": "P2", "decision": "rejected_or_out_of_scope", "resolution": "observational_audit_not_new_algorithm", "reason": "对真实健康对话做人群分层及上下文安全评分，揭示合成测试遗漏；是观察性安全审计，没有新的训练目标/模型或 Agent 核心算法，且不索取参与者对话。", "review_level": "abstract_and_bounded_original_scope_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.18465v1", "source_sha256": "1b563733d4896a6fc23c9dd754fc973a80b8aec8613790a86cac16bd20930630", "metadata_sha256": "3189980bb7552910871066009c5215a60d21da95f04de48805ec2df9e14524f4", "locations": [ "abstract", "bounded defining method / experimental setup / evaluation / limitations" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.18434", "title": "Principled Design of Expected Improvement for Continuous Simulation Optimization", "published": "2026-10-13T14:10:35Z", "revision": 2, "priority": "P2", "decision": "rejected_or_out_of_scope", "resolution": "outside_model_algorithm_scope", "reason": "异方差随机克里金/EI 的 regret 与重复分配理论，属于连续仿真优化，不是 LLM/Agent 或搜广推方法；题中 recommendation 指最终设计选择。", "review_level": "abstract_and_bounded_original_scope_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.18434v2", "source_sha256": "c6a59f910d78e7e69c17735109c5bb68163655bd028eb7615a0e0e67ae6675db", "metadata_sha256": "42d40e523cec3109f0c0f4b129e9b7d7dedd674a300dca5b13b16a06077f346a", "locations": [ "abstract", "bounded defining method / experimental setup / evaluation / limitations" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.18309", "title": "Prompt or Policy? Optimizing Physician Slate Presentations with LLM User Simulators", "published": "2026-10-13T12:52:47Z", "revision": 1, "priority": "P1", "decision": "reviewed_deferred", "resolution": "not_reproduced_high_cost", "reason": "固定医生 slate 上比较 GEPA、GRPO 及串联，使用患者模拟器而非生产转化 A/B。§2.2 为4B/12B full fine-tune、31B judge、4000 GEPA 调用及125×256 rollouts/H200；完整预算高，不复现、审查 DONE。", "review_level": "abstract_and_bounded_original_scope_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.18309v1", "source_sha256": "e842667104f8ca90428476aa0558e24a23f42b389da2fd5ef1c3fbb7e99c9052", "metadata_sha256": "d7d5b11f5470385e82fc5d6599ab10e3b58b5eb940aa9988c584ba3f638558f4", "locations": [ "§2.2 Experimental Setup" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix.", "cost_assessment": { "source_url": "https://arxiv.org/html/2610.18309v1", "basis": "§2.2 Experimental Setup", "scale": "4B/12B、31B judge、4000 metric calls、125步×256 rollouts、H200", "available_resource": "单卡 A100 或 A30，或本地 Mac；不假设多卡同时可用", "cost_type": "reported" } }, { "id": "2610.18308", "title": "SimInter: Evidence-Grounded User Simulation with Interview-Based Profile Enrichment for Conversational Recommender Systems", "published": "2026-10-13T12:52:35Z", "revision": 1, "priority": "P2", "decision": "rejected_or_out_of_scope", "resolution": "industrial_evidence_gate_not_met", "reason": "SimInter 从 ReDial 历史证据生成访谈/审计后扩充用户画像，300用户、100配对子集及双judge比较的是模拟行为多样性，不是推荐模型生产 A/B。原文 generation API 费用低，不能因此虚称高成本；按工业证据门槛不选。", "review_level": "abstract_and_bounded_original_scope_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.18308v1", "source_sha256": "f408e013fc135fa98ee90621df854c32a38d524d08c53ec48b2ef3fe5df90077", "metadata_sha256": "b79cdf8a3849c8027591013aa126dc9e5d35445fc39213113d3ea9a411257bd6", "locations": [ "abstract", "bounded defining method / experimental setup / evaluation / limitations" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.17986", "title": "Web Scale on a Budget: Tail-Aware Retrieval Evaluation", "published": "2026-10-13T08:54:15Z", "revision": 1, "priority": "P1", "decision": "reviewed_deferred", "resolution": "not_reproduced_high_cost", "reason": "每query估计 gallery 无关的检索难度，以尾部估计预测大图库指标；不能把在5M样本的估计冒称已索引100M。五backbone、多个真实大图库的原始对照仍昂贵，完整复现不做、审查 DONE；不把200GB全库描述子成本当本方法每次必需成本。", "review_level": "abstract_and_bounded_original_scope_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.17986v1", "source_sha256": "d6cd0f0ea3542af2c48d9c99052eefff7a88139805b7846a098a8565b2ef0fce", "metadata_sha256": "c1bbfa0cee95a18ee822798d8452663203e803065c4e453df6da2c7aeb6d4585", "locations": [ "Equal-cost comparison; original large-gallery validation" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix.", "cost_assessment": { "source_url": "https://arxiv.org/html/2610.17986v1", "basis": "Equal-cost comparison; original large-gallery validation", "scale": "5 backbones / 3 benchmarks / up to100M galleries；本方法5%索引仍5M", "available_resource": "单卡 A100 或 A30，或本地 Mac；不假设多卡同时可用", "cost_type": "materially_uncertain_after_read_only_checks" } }, { "id": "2610.17856", "title": "MMTA: Benchmarking Multimodal Temporal Analysis with Time Series, Text, and Vision", "published": "2026-10-13T07:20:58Z", "revision": 2, "priority": "P1", "decision": "reviewed_deferred", "resolution": "not_reproduced_high_cost", "reason": "MMTA/TimeOmni-v 将时序编码、时间对齐、视觉/文本布局和预测头结合；已核v2 Appendix C.2，使用H200 140GB、每数据集1–8卡和约40000 padded tokens/rank。完整九域三模态训练/评测成本高，不复现、审查 DONE。", "review_level": "abstract_and_bounded_original_scope_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.17856v2", "source_sha256": "9bf74f67a377e670eb292a7119ee4dd52a30c9646f38a97bdb09db21efe39bf5", "metadata_sha256": "c3bc5555a8bf31c7aa11eaee6bfb6abda9b95ed88dc3646dfceef85f780f9d79", "locations": [ "v2 Appendix C.2 Optimization" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix.", "cost_assessment": { "source_url": "https://arxiv.org/html/2610.17856v2", "basis": "v2 Appendix C.2 Optimization", "scale": "H200 140GB，1–8 GPUs/dataset，~40000 tokens/rank，九领域", "available_resource": "单卡 A100 或 A30，或本地 Mac；不假设多卡同时可用", "cost_type": "reported" } }, { "id": "2610.17841", "title": "FASHION-RQ: Mining Hierarchical Semantic Structure for Automated Evaluation of Fashion Retrieval", "published": "2026-10-13T06:59:54Z", "revision": 1, "priority": "P2", "decision": "reviewed_deferred", "resolution": "reviewed_reference_not_selected", "reason": "FASHION-RQ 用 RQ-VAE 层次生成查询及相关性、兄弟辨别/释义任务；标签由同一SID结构导出，不能当独立用户偏好真值。代码/数据在文中仍为将发布，且没有生产 A/B；记参考评测协议，不列可执行实现队列。", "review_level": "abstract_and_bounded_original_scope_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.17841v1", "source_sha256": "1fd6bed5b9f42eaa44899f7ba3c7345b8b7621167497dd55244ab5c439440c1e", "metadata_sha256": "669ec2e25ce6a591fc36cbb26331e702243d123c12123b290aa1e371f0ec8c3b", "locations": [ "abstract", "bounded defining method / experimental setup / evaluation / limitations" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.17823", "title": "Mitigating Popularity Bias in Sequential Recommendations with Sparse Prior-Regularized Objective", "published": "2026-10-13T06:43:57Z", "revision": 1, "priority": "P2", "decision": "rejected_or_out_of_scope", "resolution": "industrial_evidence_gate_not_met", "reason": "PR-entmax 将流行度先验放入稀疏概率映射，影响活跃支持和训练梯度，推理移除先验。正文§5仅公开序列数据的离线 relevance/coverage/novelty，没有生产业务 A/B，不能自行批准为经典例外。", "review_level": "abstract_and_bounded_original_scope_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.17823v1", "source_sha256": "6df1916a5eef96c210022c83b16d3b03f0a85844c62098913265cf0a06c772cb", "metadata_sha256": "53dc41fdafe3f913d19716d4763512140b9fd917ffbcf535c3d74be6fa5a83c7", "locations": [ "abstract", "bounded defining method / experimental setup / evaluation / limitations" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.17771", "title": "Many Right Answers, One Training Example: Removing Responder Noise from GRPO for LLM Recommenders", "published": "2026-10-13T05:57:20Z", "revision": 1, "priority": "P1", "decision": "reviewed_deferred", "resolution": "not_reproduced_high_cost", "reason": "ReMaRk 按请求均匀采样并对多回复标题并集计一次奖励，减少 responder 噪声。正文实报2B/4B/9B选中checkpoint仍31.5/41.6/49.3 GPU-hours，比较臂更贵；不是因为更快就便宜，不复现、审查 DONE。", "review_level": "abstract_and_bounded_original_scope_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.17771v1", "source_sha256": "4035ab977953e4ea2e3e623366004d23693712efb991abea49b828724bd5dc1b", "metadata_sha256": "8eb4dcaa31f4a5e2679c93625588bec7c5db377085e690496daff5f47ac4a601", "locations": [ "Step time and validation cost" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix.", "cost_assessment": { "source_url": "https://arxiv.org/html/2610.17771v1", "basis": "Step time and validation cost", "scale": "31.5/41.6/49.3 GPU-hours selected checkpoints + matched baseline arms", "available_resource": "单卡 A100 或 A30，或本地 Mac；不假设多卡同时可用", "cost_type": "reported" } }, { "id": "2610.17747", "title": "Auditing Failure Attribution: Label Predictability and Judge Behavior", "published": "2026-10-13T05:26:33Z", "revision": 1, "priority": "P2", "decision": "reviewed_deferred", "resolution": "reviewed_reference_protocol_not_selected", "reason": "四测试审计区分跨步骤核查与标签表面线索：输入限制、移动线索、残余准确率及聚合。保留注入/自然错误和独立judge局限，不把仅一段文本预测标签高分写成Agent推理能力；本轮只登记评测参考，不实现完整judge微调研究。", "review_level": "abstract_and_bounded_original_scope_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.17747v1", "source_sha256": "99abf29cbb528df138ed2eeb9f5492d7c9a43223bfc89c95c8b903b23adca38b", "metadata_sha256": "6ad035777558e92fcc57760ffb8e19bbd3918f914d9c394f2e8960b05cba184c", "locations": [ "abstract", "bounded defining method / experimental setup / evaluation / limitations" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.17368", "title": "HASTE: Hierarchical and Attribute-aware Supervision for Generative Recommendation", "published": "2026-10-12T20:25:50Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-12，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。HASTE: Hierarchical and Attribute-aware Supervision for Generative Recommendation。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.17368v1", "source_sha256": "4586d8836da86a759785c07e77a2a1c3c929ee522bec5c13a5cb29594596fb9d", "metadata_sha256": "0fbffb593c70ffbe8e9f26dd387773faf863e255750b80ff998e96cc3d851bb5", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.16689", "title": "EHR-Bench: A Chinese Multi-Task Benchmark for Recruitment AI", "published": "2026-10-12T12:27:33Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-12，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。EHR-Bench: A Chinese Multi-Task Benchmark for Recruitment AI。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.16689v1", "source_sha256": "bb997c4c0f9dd4b8ecc2d6ce60f754719d4406084ecdad8a452d555636746bc7", "metadata_sha256": "3f1447886acde3730e85a0c741aafd14adee36d20458c086568d6bddb07ceb6d", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.16440", "title": "ARMED: Adaptive Retrieval-Augmented Medication Recommendation", "published": "2026-10-12T09:07:30Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-12，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。ARMED: Adaptive Retrieval-Augmented Medication Recommendation。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.16440v1", "source_sha256": "088210f42484ca3ce22aa5e935dc304e6db35184b833af3d82e767bbba33d14a", "metadata_sha256": "f127ad5db0155d8bdb7f5886b9f6d0c7f33d16cfa53f160e04786621b742e213", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.16229", "title": "Tackling the Modality Inhibitance Effect with Wavelet-based Modality Integration in Multi-modal Recommendation", "published": "2026-10-12T06:05:47Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-12，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。Tackling the Modality Inhibitance Effect with Wavelet-based Modality Integration in Multi-modal Recommendation。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.16229v1", "source_sha256": "07bfa4d29366bbf1b4403805f02170f787233346b2fbe7409e3e6fe3b62b6b2c", "metadata_sha256": "ef14df573371735b977c8de3d763d498ec761ff279f778c21dd7e6e5b3f6531f", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.16038", "title": "OT-FlowRec: Optimal Transport-Guided Multi-Bridge Flow for Multimodal Sequential Recommendation", "published": "2026-10-12T01:23:07Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-12，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。OT-FlowRec: Optimal Transport-Guided Multi-Bridge Flow for Multimodal Sequential Recommendation。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.16038v1", "source_sha256": "7a77ab8fbd82ffcc00a5b2f920a1b05403ef498ee398b77b32fcd36a31c81e51", "metadata_sha256": "c03492306eb5a5e93527fb0fe2344e3176b11a4227683bc8a9336fbc766d36f8", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.15832", "title": "Local Citation Recommendation Without Graph Signals", "published": "2026-10-11T18:35:45Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-11，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。Local Citation Recommendation Without Graph Signals。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.15832v1", "source_sha256": "3352066d67ec23dbb63a53f168529af26db1601d9b8d150d3d719c7c15443d1b", "metadata_sha256": "26c1afa295c57eaf06497513d0eb93d49d9260ff3019d63f0a3d4827b2dd60f2", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.15696", "title": "DRanker: A Transformer-Based Multi-Task Ranking Model for the Disney+ Homepage", "published": "2026-10-11T15:29:51Z", "revision": 1, "priority": "P0", "decision": "reviewed_deferred", "resolution": "not_reproduced_high_cost", "reason": "DRanker 的约 7M 参数 encoder-only 多任务排序器结合因果去偏，正文有线上 A/B 与全局上线结果；不是只有摘要宣称部署。训练依赖数月 Disney 私有交互与数亿训练样本，无法在公开数据和单卡预算下忠实核验生产设置；不复现、审查 DONE。首发 arXiv 2026-10-11，属旧文补发现，不是本窗口新首发。", "review_level": "bounded_original_method_production_and_resource_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.15696v1", "source_sha256": "c62972072026bf97913645e33c61299afbc1ae89186bcdd499a3df4f6c509850", "metadata_sha256": "16f1c1336e5d072c2c3ef2352d69c9ceb1db01c24e2de2cf9a4d4c0f11f9a828", "locations": [ "§3 training data construction", "§4 model architecture", "§5.5–5.6 offline selection and online A/B", "§6 serving", "§8 global deployment" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix.", "scale": "约 7M 参数、数亿私有训练样本、频繁重训", "cost_type": "published_configuration_and_private_data_not_local_measurement", "industrial_gate_passed": true, "cost_assessment": { "source_url": "https://arxiv.org/html/2610.15696v1", "basis": "§3 training data construction; §4 model architecture; §5.5–5.6 offline selection and online A/B; §6 serving; §8 global deployment", "scale": "约 7M 参数、数亿私有训练样本、频繁重训", "available_resource": "单卡 A100 或 A30，或本地 Mac", "cost_type": "published_configuration_and_private_data_not_local_measurement" } }, { "id": "2610.15661", "title": "Feedback as an Interface: Controlling Form and Detail in LLM Translation Refinement", "published": "2026-10-11T14:39:50Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-11，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。Feedback as an Interface: Controlling Form and Detail in LLM Translation Refinement。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.15661v1", "source_sha256": "ba7b1c70c26a372924d16fa15dce42b6b31c6a29c85d6d9ad2294f975d30539b", "metadata_sha256": "212b4eb29d67618e73744dd82c128403e9d97279d708edd1e5ac657468d5adc3", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.15564", "title": "UNISON: A Unified Framework for Learning from Scored Bags", "published": "2026-10-11T12:18:17Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-11，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。UNISON: A Unified Framework for Learning from Scored Bags。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.15564v1", "source_sha256": "80e24947e32815912c3ce1eb4ef797da0bd1d119f89abdf9e1935cf3df267713", "metadata_sha256": "f3b41b805e0de453f205335a432bf577c4bed0c687e66e85540773cf54739d76", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.15523", "title": "Rethinking Evidence Modeling: Context-Coupled Graph Learning for Multimodal Recommendation", "published": "2026-10-11T10:42:40Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-11，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。Rethinking Evidence Modeling: Context-Coupled Graph Learning for Multimodal Recommendation。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.15523v1", "source_sha256": "8df11a34c12b8398885bfe5105e9723077732824e60f02c1ecace09164c3e501", "metadata_sha256": "87fd6601d11ae55adf1907cebd205df398441c53d7287699e656d22393e36305", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.15505", "title": "Federated Semantic Residuals for Multimodal Recommendation", "published": "2026-10-11T10:13:22Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-11，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。Federated Semantic Residuals for Multimodal Recommendation。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.15505v1", "source_sha256": "9231ed1ac993df130acd8e316e54cb89c42c76e0b504201c693260fd31e893a5", "metadata_sha256": "fdfd837fb08fd2e86f69840a02c801f39c0084da00b8ae3f6bebe19c319b2c04", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.15484", "title": "Toward Explainability in Multimodal Recommendation", "published": "2026-10-11T09:35:37Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-11，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。Toward Explainability in Multimodal Recommendation。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.15484v1", "source_sha256": "3fdb5c9f2044a22de82a83f4c394742df3b454742f9c9e30cbd75989654de9d7", "metadata_sha256": "eca3731499eb26c3a771f01d203f197ea2f37324efb380bcb3f96f7e166965db", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.15020", "title": "PATTERN Can Be Solved from Memory: Generator-Aware Baselines for Synthetic Node Classification", "published": "2026-10-10T15:54:55Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-10，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。PATTERN Can Be Solved from Memory: Generator-Aware Baselines for Synthetic Node Classification。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.15020v1", "source_sha256": "975a24e2e149d767559230c22c910406e45a5a9c6f30e3c030f5db788ef5b870", "metadata_sha256": "f441d756c25f02d8e6e13ad20426130d0e5ddd232dd9227b4192ed8f6fa0d1ae", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.14757", "title": "Rethinking Opinion Measurement Interfaces for Human-in-the-Loop Optimization", "published": "2026-10-10T09:25:03Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-10，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。Rethinking Opinion Measurement Interfaces for Human-in-the-Loop Optimization。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.14757v1", "source_sha256": "83842181793e6d5605088fc071723a4f6494a7a4f53fa6e4d6b41c13b0f982f6", "metadata_sha256": "97634943a701d64f57017dfa033d04118a59f64c9780098bfc2d422f9239ef0f", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.14709", "title": "Vela: User Behavior as Supervision Sparks LLM Intelligence", "published": "2026-10-10T08:28:00Z", "revision": 1, "priority": "P1", "decision": "reviewed_deferred", "resolution": "not_reproduced_high_cost", "reason": "Vela 把长期用户行为变为任务与轨迹；属于 LLM 数据/训练方案而非已证实的工业业务线上 A/B。35B-A3B、131072 长度、75K 轨迹，两 epoch；附录 RL 使用 48 GPUs（32 训练/16 rollout）。成本过高，不复现、审查 DONE。", "review_level": "bounded_original_method_production_and_resource_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.14709v1", "source_sha256": "dd866c580d5616aeecde0dc3208dd5179daf0d6922cadd81379ac6e20d2a0efc", "metadata_sha256": "72289f5f3e3ce866be2413053811efa7c60cbcf227a2d79e40764b08d0650d36", "locations": [ "§3.1 setup", "Appendix E RL configuration" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix.", "scale": "35B-A3B、131072 context、75K tasks；RL 48 GPUs", "cost_type": "published_configuration_not_local_measurement", "cost_assessment": { "source_url": "https://arxiv.org/html/2610.14709v1", "basis": "§3.1 setup; Appendix E RL configuration", "scale": "35B-A3B、131072 context、75K tasks；RL 48 GPUs", "available_resource": "单卡 A100 或 A30，或本地 Mac", "cost_type": "published_configuration_not_local_measurement" } }, { "id": "2610.14579", "title": "Fine-Tuning Strategy, Not Just Data: A Statistically Rigorous Audit of Gender and Stereotypical Bias in Embedding Models for HR Recommendation Systems", "published": "2026-10-10T04:57:38Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-10，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。Fine-Tuning Strategy, Not Just Data: A Statistically Rigorous Audit of Gender and Stereotypical Bias in Embedding Models for HR Recommendation Systems。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.14579v1", "source_sha256": "eda8bc1d64bfc61eb7f7006eeac4cc381b544725d5a787b976d31ad38004b7de", "metadata_sha256": "3adf8524925c2c2a28ffd743e14571d8d18e1f063375b8be6d13e75dc4f987ab", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.14474", "title": "RES-MR: Risk-Aware Reasoning for Explainable and Safe Medication Recommendation", "published": "2026-10-10T02:06:48Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-10，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。RES-MR: Risk-Aware Reasoning for Explainable and Safe Medication Recommendation。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.14474v1", "source_sha256": "725236ddf21ddb5970b8d03c574f980203c248c18e1ec80fc2944393720fb0f9", "metadata_sha256": "dce8a47ff933408ab5150abfcc3ae4de1185001112df7f4c5e2539acec3dcfc3", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.14394", "title": "Understanding and Mitigating Training Instability in Large-Scale Recommender Systems: A Cascade Perspective", "published": "2026-10-09T23:37:02Z", "revision": 1, "priority": "P2", "decision": "rejected_or_out_of_scope", "resolution": "industrial_evidence_gate_not_met", "reason": "Google 的 G→W→L 发散级联与 SmeLU/RMSNorm/gradient clipping、activation clipping、Huber loss 已核正文 §3；比较的是训练稳定性、AUC/RMSE，没有该方法的量化业务线上 A/B 或经批准的全流量收益。工业门槛不通过；不因 Google 署名或真实日志直接接受。", "review_level": "bounded_original_method_production_and_resource_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.14394v1", "source_sha256": "e0b6611fcd896dd6e0048a0fd7fc75b0287a995e99ab3adc8a70b06c84a52e96", "metadata_sha256": "8e3c3d8ef8c1f94760309ac92e3de7ae3728b0510437844ef428c117059c0208", "locations": [ "§2 cascade/interventions", "§3 experiments/RQ1–RQ2", "§3.1 experimental setup", "§4 conclusion" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix.", "industrial_gate_passed": false }, { "id": "2610.10076", "title": "Towards Calibrated Probabilistic Forecasts for Events of Interest via Outcome-Conditional Recalibration", "published": "2026-10-07T13:41:10Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-07，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。Towards Calibrated Probabilistic Forecasts for Events of Interest via Outcome-Conditional Recalibration。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.10076v1", "source_sha256": "c7700448f86e842405b4f34a4c4ddfbbff438e572b13bf7d07515d425cef161c", "metadata_sha256": "84c73fff4a2ee7853fcf4e1366f5ffd51941803661f81f39ff6d673000e556be", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.08652", "title": "Steering Diffusion Models to Rare Events with Sequential Monte Carlo", "published": "2026-10-06T16:38:03Z", "revision": 2, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-06，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。Steering Diffusion Models to Rare Events with Sequential Monte Carlo。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.08652v1", "source_sha256": "a2f30b9c24e187eb73b9ee8c2d6e81c3eb9cdc551460b425c448e3737ea1d72e", "metadata_sha256": "3073fdb3744361dfa21f4abbb49c2c2b962afd69db0fadf95d5ed64ee02f9bff", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.04444", "title": "Causally Fair Generation with Large Language Models", "published": "2026-10-03T11:14:12Z", "revision": 1, "priority": "P2", "decision": "historical", "resolution": "historical_identity_scope_triage", "reason": "首发 2026-10-03，早于本轮窗口；本轮按身份/日期与摘要范围分流，不计新首发或完整算法审查。Causally Fair Generation with Large Language Models。", "review_level": "identity_date_and_abstract_scope", "abstract_review_completed": true, "fulltext_review_completed": false, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.04444v1", "source_sha256": "300c73972fea32d96137a14805576cb619946fca90550d96a26ae20eb512fbb3", "metadata_sha256": "14ec2f362933c1aac9669db76fdd4d24c37482cca0b52e78d91a65c65cec7a34", "locations": [ "arXiv API published + abstract" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix." }, { "id": "2610.13989", "title": "PsyBer-Agent: A Psychology-Driven User Agent for Recommender Systems", "published": "2026-09-26T14:42:02Z", "revision": 1, "priority": "P2", "decision": "reviewed_deferred", "resolution": "not_reproduced_high_cost", "reason": "PsyBer-Agent 的三状态 Psy-Engine、UGW 对齐与 WebSim 是用户模拟。§5.2 虽称 A/B，实际固定后端比较 500 人与 500 模拟用户的行为一致性，不是推荐器生产业务 A/B；不能通过工业门槛。原评测使用 Qwen3-VL-235B 等大量闭环模拟，成本过高不复现，审查 DONE。", "review_level": "bounded_original_method_production_and_resource_evidence", "abstract_review_completed": true, "fulltext_review_completed": true, "full_article_read": false, "source_url": "https://arxiv.org/html/2610.13989v1", "source_sha256": "9b69089f30900262727eef42aa3c9bd39f0ea61d6663fc9ffafc4737c831a31c", "metadata_sha256": "1394ac4c2099275704912d0b3a5081cf5dde5dc434e099613625f29df094c81e", "locations": [ "§3 Psy-Engine and alignment", "§5.2 fixed-backend agreement", "Appendix D.6", "Appendix E.9" ], "review_done": true, "automatic_retry": false, "automatic_implementation_queue": false, "implementation_status": "not_implemented_by_review", "read_scope": "Only recorded scope/evaluation/resource passages; not every main section or appendix.", "scale": "235B VLM、长闭环和大规模模拟", "cost_type": "published_configuration_not_local_measurement", "industrial_gate_passed": false, "cost_assessment": { "source_url": "https://arxiv.org/html/2610.13989v1", "basis": "§3 Psy-Engine and alignment; §5.2 fixed-backend agreement; Appendix D.6; Appendix E.9", "scale": "235B VLM、长闭环和大规模模拟", "available_resource": "单卡 A100 或 A30，或本地 Mac", "cost_type": "published_configuration_not_local_measurement" } } ] }