{ "schema_version": 1, "batches": [ { "batch": "2026-08-08-p0-p1-closed-audit", "checked_at": "2026-08-08T23:59:59+08:00", "window": "2026-01-01 through 2026-08-08", "sources": [ "arXiv", "institution reverse search", "citation snowball", "official repositories" ], "required_tracks": [ "recommendation", "foundation-model", "post-training", "agent" ], "candidates": [ { "id": "2608.02148", "priority": "P0", "track": "recommendation", "status": "implemented", "key": "dme", "doc": "reproductions/2608.02148-dme/README.md" }, { "id": "2608.01949", "priority": "P0", "track": "recommendation", "status": "implemented", "key": "steps", "doc": "reproductions/2608.01949-steps/README.md" }, { "id": "2608.01738", "priority": "P0", "track": "recommendation", "status": "implemented", "key": "spear", "doc": "reproductions/2608.01738-spear/README.md" }, { "id": "2607.17247", "priority": "P0", "track": "post-training", "status": "implemented", "key": "distilled-rl", "doc": "post-training/2607.17247-distilled-rl/README.md" }, { "id": "2608.06296", "priority": "P0", "track": "post-training", "status": "implemented", "key": "u-opsd", "doc": "post-training/2608.06296-u-opsd/README.md" }, { "id": "2608.05987", "priority": "P0", "track": "agent", "status": "implemented", "key": "agent-opsd", "doc": "agent-research/2608.05987-agent-opsd/README.md" }, { "id": "2608.04788", "priority": "P0", "track": "agent", "status": "implemented", "key": "ocsd", "doc": "agent-research/2608.04788-ocsd/README.md" }, { "id": "2608.03137", "priority": "P0", "track": "agent", "status": "implemented", "key": "vermem", "doc": "agent-research/2608.03137-vermem/README.md" }, { "id": "2608.06347", "priority": "P1", "track": "post-training", "status": "implemented", "key": "rp-opsd", "doc": "post-training/2608.06347-rp-opsd/README.md" }, { "id": "2608.01837", "priority": "P1", "track": "post-training", "status": "implemented", "key": "pcsd", "doc": "post-training/2608.01837-pcsd/README.md" }, { "id": "2608.03223", "priority": "P1", "track": "post-training", "status": "implemented", "key": "adrs", "doc": "post-training/2608.03223-adrs/README.md" }, { "id": "2608.01739", "priority": "P1", "track": "agent", "status": "implemented", "key": "coevo-mem", "doc": "agent-research/2608.01739-coevo-mem/README.md" }, { "id": "2606.30406", "priority": "P1", "track": "post-training", "status": "implemented", "key": "mopd", "doc": "post-training/2606.30406-mopd/README.md" }, { "id": "2606.06712", "priority": "P1", "track": "post-training", "status": "implemented", "key": "opd-lm", "doc": "post-training/2606.06712-opd-lm/README.md" }, { "id": "2607.16669", "priority": "P1", "track": "foundation-model", "status": "implemented", "key": "open-language-model", "doc": "reproductions/2607.16669-open-language-model/README.md" } ] }, { "batch": "2026-08-08-global-theme-gap-review", "scope_kind": "global", "checked_at": "2026-08-08T23:59:59+08:00", "window": "classics through 2026-08-08", "sources": [ "arXiv", "ACL Anthology", "official project pages", "official repositories", "institution reverse search", "citation snowball" ], "required_tracks": [ "recommendation", "foundation-model", "post-training", "agent" ], "required_subtopics": [ { "track": "recommendation", "subtopic": "retrieval" }, { "track": "recommendation", "subtopic": "pre-ranking" }, { "track": "recommendation", "subtopic": "ranking" }, { "track": "recommendation", "subtopic": "reranking-mixing" }, { "track": "recommendation", "subtopic": "generative-recommendation" }, { "track": "recommendation", "subtopic": "content-understanding" }, { "track": "recommendation", "subtopic": "training-serving" }, { "track": "foundation-model", "subtopic": "architecture" }, { "track": "foundation-model", "subtopic": "position-long-context" }, { "track": "foundation-model", "subtopic": "pretraining-data" }, { "track": "foundation-model", "subtopic": "multimodal" }, { "track": "foundation-model", "subtopic": "inference-serving" }, { "track": "post-training", "subtopic": "classic-rlhf-preference" }, { "track": "post-training", "subtopic": "reward-process-supervision" }, { "track": "post-training", "subtopic": "reasoning-rl" }, { "track": "post-training", "subtopic": "off-policy-stability" }, { "track": "post-training", "subtopic": "self-play-distillation" }, { "track": "agent", "subtopic": "planning-reasoning" }, { "track": "agent", "subtopic": "tool-use" }, { "track": "agent", "subtopic": "memory-skills" }, { "track": "agent", "subtopic": "agentic-rl" }, { "track": "agent", "subtopic": "multi-agent" }, { "track": "agent", "subtopic": "benchmarks-environments" } ], "candidates": [ { "id": "2604.25291", "title": "GloRank", "priority": "P0", "track": "recommendation", "subtopic": "generative-recommendation", "status": "implemented", "key": "glorank", "doc": "reproductions/2604.25291-glorank/README.md" }, { "id": "2604.07420", "title": "Dual-Rerank", "priority": "P0", "track": "recommendation", "subtopic": "reranking-mixing", "status": "implemented", "key": "dual-rerank", "doc": "reproductions/2604.07420-dual-rerank/README.md" }, { "id": "2603.02999", "title": "OneRanker", "priority": "P0", "track": "recommendation", "subtopic": "ranking", "status": "implemented", "key": "oneranker", "doc": "reproductions/2603.02999-oneranker/README.md" }, { "id": "2506.07261", "title": "RADAR", "priority": "P0", "track": "recommendation", "subtopic": "retrieval", "status": "implemented", "key": "radar", "doc": "reproductions/2506.07261-radar/README.md" }, { "id": "2511.12518", "title": "DualGR", "priority": "P0", "track": "recommendation", "subtopic": "retrieval", "status": "implemented", "key": "dualgr", "doc": "reproductions/2511.12518-dualgr/README.md" }, { "id": "2508.20400", "title": "MPFormer", "priority": "P0", "track": "recommendation", "subtopic": "training-serving", "status": "implemented", "key": "mpformer", "doc": "reproductions/2508.20400-mpformer/README.md" }, { "id": "2603.03770", "title": "HAP", "priority": "P0", "track": "recommendation", "subtopic": "pre-ranking", "status": "implemented", "key": "hap", "doc": "reproductions/2603.03770-hap/README.md" }, { "id": "2509.18091", "title": "OnePiece", "priority": "P0", "track": "recommendation", "subtopic": "generative-recommendation", "status": "implemented", "key": "onepiece", "doc": "reproductions/2509.18091-onepiece/README.md" }, { "id": "2509.21179", "title": "IntSR", "priority": "P0", "track": "recommendation", "subtopic": "content-understanding", "status": "implemented", "key": "intsr", "doc": "reproductions/2509.21179-intsr/README.md" }, { "id": "2407.16357", "title": "TWIN-V2", "priority": "P1", "track": "recommendation", "subtopic": "training-serving", "status": "implemented", "key": "twin-v2", "doc": "reproductions/2407.16357-twin-v2/README.md" }, { "id": "2406.09021", "title": "CDM", "priority": "P0", "track": "recommendation", "subtopic": "reranking-mixing", "status": "implemented", "key": "cdm", "doc": "reproductions/2406.09021-cdm/README.md" }, { "id": "2406.07932", "title": "CWM", "priority": "P0", "track": "recommendation", "subtopic": "ranking", "status": "implemented", "key": "cwm", "doc": "reproductions/2406.07932-cwm/README.md" }, { "id": "2006.05639", "title": "SIM", "priority": "P1", "track": "recommendation", "subtopic": "ranking", "status": "implemented", "key": "sim", "doc": "reproductions/2006.05639-sim/README.md" }, { "id": "2510.11056", "title": "CRSD", "priority": "P1", "track": "recommendation", "subtopic": "content-understanding", "status": "implemented", "key": "crsd", "doc": "reproductions/2510.11056-crsd/README.md" }, { "id": "2104.09864", "title": "RoFormer / RoPE", "priority": "P0", "track": "foundation-model", "subtopic": "position-long-context", "status": "implemented", "key": "rope", "doc": "reproductions/2104.09864-rope/README.md" }, { "id": "2108.12409", "title": "ALiBi", "priority": "P0", "track": "foundation-model", "subtopic": "position-long-context", "status": "implemented", "key": "alibi", "doc": "reproductions/2108.12409-alibi/README.md" }, { "id": "2305.13245", "title": "GQA", "priority": "P0", "track": "foundation-model", "subtopic": "architecture", "status": "implemented", "key": "gqa", "doc": "reproductions/2305.13245-gqa/README.md" }, { "id": "2411.13676", "title": "Hymba", "priority": "P0", "track": "foundation-model", "subtopic": "architecture", "status": "implemented", "key": "hymba", "doc": "reproductions/2411.13676-hymba/README.md" }, { "id": "2502.13189", "title": "MoBA", "priority": "P0", "track": "foundation-model", "subtopic": "architecture", "status": "implemented", "key": "moba", "doc": "reproductions/2502.13189-moba/README.md" }, { "id": "2305.10429", "title": "DoReMi", "priority": "P0", "track": "foundation-model", "subtopic": "pretraining-data", "status": "implemented", "key": "doremi", "doc": "reproductions/2305.10429-doremi/README.md" }, { "id": "2403.16952", "title": "Data Mixing Laws", "priority": "P0", "track": "foundation-model", "subtopic": "pretraining-data", "status": "implemented", "key": "data-mixing-laws", "doc": "reproductions/2403.16952-data-mixing-laws/README.md" }, { "id": "2412.09871", "title": "Byte Latent Transformer", "priority": "P0", "track": "foundation-model", "subtopic": "architecture", "status": "implemented", "key": "blt", "doc": "reproductions/2412.09871-blt/README.md" }, { "id": "2103.00020", "title": "CLIP", "priority": "P1", "track": "foundation-model", "subtopic": "multimodal", "status": "implemented", "key": "clip", "doc": "reproductions/2103.00020-clip/README.md" }, { "id": "2304.08485", "title": "LLaVA", "priority": "P1", "track": "foundation-model", "subtopic": "multimodal", "status": "implemented", "key": "llava", "doc": "reproductions/2304.08485-llava/README.md" }, { "id": "2211.17192", "title": "Speculative Decoding", "priority": "P1", "track": "foundation-model", "subtopic": "inference-serving", "status": "implemented", "key": "speculative-decoding", "doc": "reproductions/2211.17192-speculative-decoding/README.md" }, { "id": "2306.00978", "title": "AWQ", "priority": "P1", "track": "foundation-model", "subtopic": "inference-serving", "status": "implemented", "key": "awq", "doc": "reproductions/2306.00978-awq/README.md" }, { "id": "2401.10774", "title": "Medusa", "priority": "P1", "track": "foundation-model", "subtopic": "inference-serving", "status": "implemented", "key": "medusa", "doc": "reproductions/2401.10774-medusa/README.md" }, { "id": "2309.00267", "title": "RLAIF", "priority": "P0", "track": "post-training", "subtopic": "classic-rlhf-preference", "status": "implemented", "key": "rlaif", "doc": "post-training/2309.00267-rlaif/README.md" }, { "id": "2305.20050", "title": "Let's Verify Step by Step", "priority": "P0", "track": "post-training", "subtopic": "reward-process-supervision", "status": "implemented", "key": "process-supervision", "doc": "post-training/2305.20050-process-supervision/README.md" }, { "id": "2312.08935", "title": "Math-Shepherd", "priority": "P0", "track": "post-training", "subtopic": "reward-process-supervision", "status": "implemented", "key": "math-shepherd", "doc": "post-training/2312.08935-math-shepherd/README.md" }, { "id": "2401.10020", "title": "Self-Rewarding Language Models", "priority": "P0", "track": "post-training", "subtopic": "self-play-distillation", "status": "implemented", "key": "self-rewarding", "doc": "post-training/2401.10020-self-rewarding/README.md" }, { "id": "2504.14945", "title": "LUFFY", "priority": "P0", "track": "post-training", "subtopic": "reasoning-rl", "status": "implemented", "key": "luffy", "doc": "post-training/2504.14945-luffy/README.md" }, { "id": "2504.16084", "title": "TTRL", "priority": "P0", "track": "post-training", "subtopic": "reasoning-rl", "status": "implemented", "key": "ttrl", "doc": "post-training/2504.16084-ttrl/README.md" }, { "id": "2505.03335", "title": "Absolute Zero", "priority": "P0", "track": "post-training", "subtopic": "self-play-distillation", "status": "implemented", "key": "absolute-zero", "doc": "post-training/2505.03335-absolute-zero/README.md" }, { "id": "2505.19590", "title": "INTUITOR", "priority": "P0", "track": "post-training", "subtopic": "reasoning-rl", "status": "implemented", "key": "intuitor", "doc": "post-training/2505.19590-intuitor/README.md" }, { "id": "2506.13585", "title": "CISPO", "priority": "P0", "track": "post-training", "subtopic": "off-policy-stability", "status": "implemented", "key": "cispo", "doc": "post-training/2506.13585-cispo/README.md" }, { "id": "2506.24119", "title": "SPIRAL", "priority": "P0", "track": "post-training", "subtopic": "self-play-distillation", "status": "implemented", "key": "spiral", "doc": "post-training/2506.24119-spiral/README.md" }, { "id": "2605.12969", "title": "ConSPO", "priority": "P0", "track": "post-training", "subtopic": "off-policy-stability", "status": "implemented", "key": "conspo", "doc": "post-training/2605.12969-conspo/README.md" }, { "id": "2512.01374", "title": "MiniRL", "priority": "P1", "track": "post-training", "subtopic": "reasoning-rl", "status": "implemented", "key": "minirl", "doc": "post-training/2512.01374-minirl/README.md" }, { "id": "2605.12070", "title": "Missing Old Logits", "priority": "P1", "track": "post-training", "subtopic": "off-policy-stability", "status": "implemented", "key": "missing-old-logits", "doc": "post-training/2605.12070-missing-old-logits/README.md" }, { "id": "2606.19236", "title": "STARE", "priority": "P1", "track": "post-training", "subtopic": "reward-process-supervision", "status": "implemented", "key": "stare", "doc": "post-training/2606.19236-stare/README.md" }, { "id": "2504.03160", "title": "DeepResearcher", "priority": "P0", "track": "agent", "subtopic": "planning-reasoning", "status": "implemented", "key": "deepresearcher", "doc": "agent-research/2504.03160-deepresearcher/README.md" }, { "id": "2504.11536", "title": "ReTool", "priority": "P0", "track": "agent", "subtopic": "tool-use", "status": "implemented", "key": "retool", "doc": "agent-research/2504.11536-retool/README.md" }, { "id": "2504.13958", "title": "ToolRL", "priority": "P0", "track": "agent", "subtopic": "tool-use", "status": "implemented", "key": "toolrl", "doc": "agent-research/2504.13958-toolrl/README.md" }, { "id": "2512.17102", "title": "SAGE", "priority": "P0", "track": "agent", "subtopic": "memory-skills", "status": "implemented", "key": "sage", "doc": "agent-research/2512.17102-sage/README.md" }, { "id": "2602.02474", "title": "MemSkill", "priority": "P0", "track": "agent", "subtopic": "memory-skills", "status": "implemented", "key": "memskill", "doc": "agent-research/2602.02474-memskill/README.md" }, { "id": "2603.18743", "title": "Memento-Skills", "priority": "P0", "track": "agent", "subtopic": "memory-skills", "status": "implemented", "key": "memento-skills", "doc": "agent-research/2603.18743-memento-skills/README.md" }, { "id": "2604.07791", "title": "SEARL", "priority": "P0", "track": "agent", "subtopic": "agentic-rl", "status": "implemented", "key": "searl", "doc": "agent-research/2604.07791-searl/README.md" }, { "id": "2511.16043", "title": "Agent0", "priority": "P0", "track": "agent", "subtopic": "multi-agent", "status": "implemented", "key": "agent0", "doc": "agent-research/2511.16043-agent0/README.md" }, { "id": "2511.14460", "title": "Agent-R1", "priority": "P1", "track": "agent", "subtopic": "agentic-rl", "status": "implemented", "key": "agent-r1", "doc": "agent-research/2511.14460-agent-r1/README.md" }, { "id": "2303.17760", "title": "CAMEL", "priority": "P1", "track": "agent", "subtopic": "multi-agent", "status": "implemented", "key": "camel", "doc": "agent-research/2303.17760-camel/README.md" }, { "id": "2305.16504", "title": "ToolBench / ToolLLM", "priority": "P1", "track": "agent", "subtopic": "tool-use", "status": "implemented", "key": "toolbench", "doc": "agent-research/2305.16504-toolbench/README.md" }, { "id": "2311.12983", "title": "GAIA", "priority": "P1", "track": "agent", "subtopic": "benchmarks-environments", "status": "implemented", "key": "gaia", "doc": "agent-research/2311.12983-gaia/README.md" } ] }, { "batch": "2026-08-09-google-meta-priority-correction", "scope_kind": "institution-priority", "checked_at": "2026-08-09T23:59:59+08:00", "window": "2026-01-01 through 2026-08-09", "sources": [ "official arXiv API", "paper PDF affiliation and evidence review", "GitHub public repository search" ], "required_tracks": [ "recommendation" ], "priority_institutions": [ "Google", "Meta" ], "institution_sweeps": [ { "organization": "Google", "queries": [ "Google DeepMind / YouTube author affiliations", "YouTube recommendation live launch / full traffic / online A/B" ], "candidate_ids": [ "2606.25147", "2607.27577", "2607.24865", "2604.26969" ], "candidate_discovery_gate": "affiliation-and-topic", "abstract_online_evidence_required": false, "full_text_review_required": true }, { "organization": "Meta", "queries": [ "Meta Platforms author affiliations", "Instagram recommendation deployed / full traffic / online A/B" ], "candidate_ids": [ "2607.12281", "2602.16124", "2602.10016", "2602.16986" ], "candidate_discovery_gate": "affiliation-and-topic", "abstract_online_evidence_required": false, "full_text_review_required": true } ], "candidates": [ { "id": "2606.25147", "title": "TokenMinds", "organization": "Google DeepMind / YouTube", "evidence_gate": "quantified-online-ab-and-full-traffic", "priority": "P0", "priority_reason": "Google/YouTube industrial recommendation with seven-day A/B and full-user-traffic deployment", "evidence_review": { "scope": "full-text", "abstract_used_as_gate": false, "locations": [ "Section 4.3", "Table 4", "Table 6" ], "matched_terms": [ "A/B test", "seven-day experiments", "full user traffic" ] }, "track": "recommendation", "status": "implemented", "key": "tokenminds", "doc": "reproductions/2606.25147-tokenminds/README.md" }, { "id": "2607.12281", "title": "SlimPer", "organization": "Meta Platforms, Inc.", "evidence_gate": "statistically-significant-full-traffic", "priority": "P0", "priority_reason": "Meta industrial recommendation deployed at full traffic on Instagram Reels and Feed", "evidence_review": { "scope": "full-text", "abstract_used_as_gate": false, "locations": [ "Section 4.3 A/B Test Performance", "Section 6 Conclusion" ], "matched_terms": [ "launched to full traffic", "statistically significant gains", "controlled A/B experiments" ] }, "track": "recommendation", "status": "implemented", "key": "slimper", "doc": "reproductions/2607.12281-slimper/README.md" }, { "id": "2607.27577", "title": "Heterogeneous Ranking in Industrial-Scale Recommender Systems: A Case Study", "organization": "Google / Discover", "evidence_gate": "quantified-online-ab", "priority": "P0", "priority_reason": "Google Discover heterogeneous ranking with 7-day 1% live-traffic A/B", "evidence_review": { "scope": "full-text", "abstract_used_as_gate": false, "locations": [ "Section 4.4", "Table 4" ], "matched_terms": [ "live experiments", "1% traffic", "7 days" ] }, "track": "recommendation", "status": "implemented", "key": "ha-moe", "doc": "reproductions/2607.27577-ha-moe/README.md" }, { "id": "2607.24865", "title": "Tokens are All You Need: Dual-purpose Semantic IDs for Achieving LLM-Level I/O Efficiency in Recommendation Systems", "organization": "Google DeepMind / YouTube", "evidence_gate": "quantified-production-online", "priority": "P0", "priority_reason": "YouTube ranking and retrieval production deployment with quantified online gains", "evidence_review": { "scope": "full-text", "abstract_used_as_gate": false, "locations": [ "Section 4.2", "Table 1" ], "matched_terms": [ "online experiments", "production", "ranking", "retrieval" ] }, "track": "recommendation", "status": "implemented", "key": "dual-sid", "doc": "reproductions/2607.24865-dual-sid/README.md" }, { "id": "2604.26969", "title": "AgenticRecTune: Multi-Agent with Self-Evolving Skillhub for Recommendation System Optimization", "organization": "Google / Discover", "evidence_gate": "quantified-online-ab", "priority": "P0", "priority_reason": "Google Discover agentic optimization with statistically significant live-traffic A/B", "evidence_review": { "scope": "full-text", "abstract_used_as_gate": false, "locations": [ "Section 5", "Table 1", "Table 3" ], "matched_terms": [ "online A/B testing", "live traffic", "p < 0.05" ] }, "track": "recommendation", "status": "implemented", "key": "agentic-rec-tune", "doc": "reproductions/2604.26969-agentic-rec-tune/README.md" }, { "id": "2602.16124", "title": "Rethinking ANN-based Retrieval: Multifaceted Learnable Index for Large-scale Recommendation System", "organization": "Meta", "evidence_gate": "quantified-online-ab", "priority": "P0", "priority_reason": "Meta retrieval model with 7-day production A/B and QPS gain", "evidence_review": { "scope": "full-text", "abstract_used_as_gate": false, "locations": [ "Section 4.6" ], "matched_terms": [ "online A/B test", "7 days", "QPS" ] }, "track": "recommendation", "status": "implemented", "key": "mfli", "doc": "reproductions/2602.16124-mfli/README.md" }, { "id": "2602.10016", "title": "Kunlun: Establishing Scaling Laws for Massive-Scale Recommendation Systems through Unified Architecture Design", "organization": "Meta", "evidence_gate": "quantified-full-deployment", "priority": "P0", "priority_reason": "Deployed across major Meta Ads models with quantified topline gain", "evidence_review": { "scope": "full-text", "abstract_used_as_gate": false, "locations": [ "Section 5.5" ], "matched_terms": [ "deployed", "major Ads models", "topline" ] }, "track": "recommendation", "status": "implemented", "key": "kunlun", "doc": "reproductions/2602.10016-kunlun/README.md" }, { "id": "2602.16986", "title": "Bending the Scaling Law Curve in Large-Scale Recommendation Systems", "organization": "Meta", "evidence_gate": "quantified-online-ab-and-full-deployment", "priority": "P0", "priority_reason": "Meta recommendation with 30-day A/B and full deployment to billions of daily users", "evidence_review": { "scope": "full-text", "abstract_used_as_gate": false, "locations": [ "Section 5.4" ], "matched_terms": [ "30-day A/B", "fully deployed", "billions of daily active users" ] }, "track": "recommendation", "status": "implemented", "key": "ultra-hstu", "doc": "reproductions/2602.16986-ultra-hstu/README.md" } ] }, { "batch": "2026-08-13-latest-cross-domain-review", "scope_kind": "global", "checked_at": "2026-08-13T12:30:00+08:00", "window": "2026-08-10 through 2026-08-13", "sources": [ "official arXiv API", "paper PDF full-text review", "Google and Meta affiliation reverse search", "official repository search" ], "required_tracks": [ "recommendation", "foundation-model", "post-training", "agent" ], "required_subtopics": [ { "track": "recommendation", "subtopic": "generative-ranking" }, { "track": "foundation-model", "subtopic": "multimodal" }, { "track": "post-training", "subtopic": "preference-optimization" }, { "track": "agent", "subtopic": "agentic-rl" } ], "candidates": [ { "id": "2608.09440", "title": "MetaStrategy", "priority": "P0", "track": "recommendation", "subtopic": "generative-ranking", "status": "implemented", "key": "metastrategy", "doc": "reproductions/2608.09440-metastrategy/README.md" }, { "id": "2608.11015", "title": "Sona Technical Report", "priority": "P0", "track": "recommendation", "subtopic": "generative-ranking", "status": "implemented", "key": "sona", "doc": "reproductions/2608.11015-sona/README.md" }, { "id": "2608.12209", "title": "Generation as Auxiliary Supervision", "priority": "P0", "track": "foundation-model", "subtopic": "multimodal", "status": "implemented", "key": "gas", "doc": "reproductions/2608.12209-gas/README.md" }, { "id": "2608.12062", "title": "Preference Tree Optimization", "priority": "P0", "track": "post-training", "subtopic": "preference-optimization", "status": "implemented", "key": "pto", "doc": "post-training/2608.12062-pto/README.md" }, { "id": "2608.12158", "title": "C2-DPO", "priority": "P0", "track": "post-training", "subtopic": "preference-optimization", "status": "implemented", "key": "c2-dpo", "doc": "post-training/2608.12158-c2-dpo/README.md" }, { "id": "2608.10357", "title": "Efficient Reinforcement Learning for Long-Horizon Tool-Use Agentic Tasks", "priority": "P0", "track": "agent", "subtopic": "agentic-rl", "status": "implemented", "key": "sinkflex-rl", "doc": "agent-research/2608.10357-sinkflex-rl/README.md" }, { "id": "2608.12184", "title": "GALLM", "priority": "P1", "track": "recommendation", "subtopic": "generative-ranking", "status": "rejected", "reason": "industrial recommendation paper reports offline evaluation but no quantified online A/B or confirmed full-traffic deployment" }, { "id": "2608.12149", "title": "Massive Activations in Large Language Models", "priority": "P1", "track": "foundation-model", "subtopic": "multimodal", "status": "rejected", "reason": "analysis paper does not introduce an independently executable architecture or training algorithm for the reproduction/evolve queue" } ] }, { "batch": "2026-08-13-genrec-high-recall-correction", "scope_kind": "high-recall-correction", "checked_at": "2026-08-13T19:40:00+08:00", "window": "2026-08-10 through 2026-08-13", "sources": [ "official arXiv API multi-query pagination", "paper PDF full-text review", "priority organization reverse search", "GitHub public repository search" ], "query_matrix": [ "recsys-general", "recommendation-ranking", "generative-recommendation", "llm-recommendation", "industrial-ranking", "production-evidence", "deployment-evidence", "search-ranking", "advertising-ranking" ], "pagination": { "page_size": 50, "maximum_results_per_query": 200, "deduplicate_by": "canonical arXiv ID" }, "priority_organization_terms": [ "Google", "Google DeepMind", "Meta", "Netflix", "ByteDance", "TikTok", "Alibaba", "Taobao", "Kuaishou", "Pinterest", "Amazon", "Microsoft", "Tencent" ], "organization_query_matrix": [ "priority-org-google", "priority-org-google-deepmind", "priority-org-meta", "priority-org-netflix", "priority-org-bytedance", "priority-org-tiktok", "priority-org-alibaba", "priority-org-taobao", "priority-org-kuaishou", "priority-org-pinterest", "priority-org-amazon", "priority-org-microsoft", "priority-org-tencent" ], "failure_corrected": "The previous global audit treated one candidate per declared subtopic as coverage and used a narrow top-result query. It therefore missed GenRec even though it was inside the audited date window.", "candidates": [ { "id": "2608.10257", "title": "GenRec: An LLM-Backed Recommendation Ranker at Netflix", "priority": "P0", "track": "recommendation", "subtopic": "llm-ranking-post-training", "organization": "Netflix", "matched_queries": [ "llm-recommendation", "priority-org-netflix", "recsys-general", "recommendation-ranking", "production-evidence" ], "status": "implemented", "key": "genrec-netflix", "doc": "reproductions/2608.10257-genrec-netflix/README.md", "evidence_gate": "quantified large-scale online A/B", "priority_reason": "Netflix production LLM ranker with four-week, approximately 10% traffic A/B and statistically significant short- and long-term metrics", "evidence_review": { "scope": "full-text", "abstract_used_as_gate": false, "locations": [ "Section 5.1", "Figure 3", "Section 4.5", "Section 4.6" ], "matched_terms": [ "approximately 10% of Netflix traffic", "4 weeks", "+0.115%", "+0.006%", "P = 3.1e-10", "P = 0.025" ] } } ] }, { "batch": "2026-08-24-recent-representative-review", "scope_kind": "recent-representative-review", "checked_at": "2026-08-24T18:00:00+08:00", "window": "2026-08-10 through 2026-08-24", "sources": [ "official arXiv API multi-query pagination", "cross-source official publication pages", "paper HTML/PDF full-text review", "Google and Meta priority reverse search", "official repository search" ], "candidate_pool": { "recommendation": { "total": 74, "new": 68 }, "foundation-model": { "total": 505, "new": 502 }, "post-training": { "total": 162, "new": 161 }, "agent": { "total": 430, "new": 428 } }, "closure_claim": "representative full-text shortlist only; the high-recall candidate artifacts remain pending incremental review", "candidates": [ { "id": "2608.18606", "title": "OneModel: A Unified Foundation for Platform-Scale Multi-Scenario Ranking", "organization": "Xiaohongshu", "priority": "P0", "track": "recommendation", "subtopic": "multi-scenario-ranking", "status": "implemented", "key": "onemodel", "doc": "reproductions/2608.18606-onemodel/README.md", "evidence_gate": "three quantified production A/B experiments", "priority_reason": "platform-scale ranking foundation with online recommendation, advertising and merchant results" }, { "id": "2608.21236", "title": "RARE: Decoupling Representation Steering from Expert Routing in Mixture-of-Experts Language Models", "organization": "Huazhong University of Science and Technology", "priority": "P0", "track": "foundation-model", "subtopic": "moe-routing-and-steering", "status": "implemented", "key": "rare", "doc": "reproductions/2608.21236-rare/README.md" }, { "id": "2608.11674", "title": "GCPO: Diagnosing and Constraining Subspace Geometry in Rollout RL for LLMs", "organization": "Shanghai AI Laboratory", "priority": "P0", "track": "post-training", "subtopic": "geometric-constrained-policy-optimization", "status": "implemented", "key": "gcpo", "doc": "post-training/2608.11674-gcpo/README.md" }, { "id": "2608.21292", "title": "AUSO: Action-Level Unified Skill Optimization from Internalization to Utilization", "organization": "University of Science and Technology of China", "priority": "P0", "track": "agent", "subtopic": "action-level-skill-optimization", "status": "implemented", "key": "auso", "doc": "agent-research/2608.21292-auso/README.md" }, { "id": "2608.18531", "title": "Pairwise Ranking", "organization": "not a qualifying Google/Meta industrial recommendation paper", "priority": "P2", "track": "recommendation", "status": "rejected", "reason": "priority-organization query false positive; topic and full-text scope do not satisfy the industrial recommendation online-evidence gate" }, { "id": "2608.15780", "title": "Decomposing Staleness", "organization": "not a qualifying Google/Meta industrial recommendation paper", "priority": "P2", "track": "recommendation", "status": "rejected", "reason": "priority-organization query false positive; outside the repository's industrial recommendation implementation scope" }, { "id": "2608.15424", "title": "ETHOS", "organization": "not a qualifying Google/Meta industrial recommendation paper", "priority": "P2", "track": "recommendation", "status": "rejected", "reason": "priority-organization query false positive; no qualifying industrial recommendation online-evidence target was found" } ] }, { "batch": "2026-08-25-daily-increment", "scope_kind": "daily-high-recall-increment", "checked_at": "2026-08-25T12:00:00+08:00", "window": "new arXiv daily batch visible on 2026-08-25; submitted 2026-08-24 UTC", "sources": [ "official arXiv API submittedDate exhaustive metadata", "DeepXiv semantic cross-check", "paper HTML/PDF full-text review", "Google and Meta priority reverse search", "official repository links" ], "candidate_pool": { "cross_domain_metadata_total": 329 }, "industrial_recommendation_result": "no new paper passed the quantified online A/B or full-traffic evidence gate", "candidates": [ { "id": "2608.23493", "key": "srpo", "track": "post-training", "priority": "P0", "status": "implemented", "doc": "post-training/2608.23493-srpo/README.md" }, { "id": "2608.23311", "key": "erpo", "track": "post-training", "priority": "P0", "organization": "AMAP, Alibaba Group", "status": "implemented", "doc": "post-training/2608.23311-erpo/README.md" }, { "id": "2608.23318", "key": "agent-g2", "track": "agent", "priority": "P0", "organization": "Baidu", "status": "implemented", "doc": "agent-research/2608.23318-agent-g2/README.md" }, { "id": "2608.23041", "key": "autosaddler", "track": "agent", "priority": "P0", "organization": "Microsoft / POSTECH", "status": "implemented", "doc": "agent-research/2608.23041-autosaddler/README.md" }, { "id": "2608.23400", "track": "recommendation", "priority": "P2", "status": "rejected", "reason": "public-dataset academic recommendation; no quantified production online A/B" }, { "id": "2608.22734", "track": "recommendation", "priority": "P2", "status": "rejected", "reason": "public-dataset academic recommendation; no quantified production online A/B" } ] }, { "batch": "2026-08-26-daily-increment", "scope_kind": "daily-high-recall-increment", "checked_at": "2026-08-26T18:00:00+08:00", "window": "new arXiv listings visible on 2026-08-26; submitted 2026-08-25 UTC", "sources": [ "official arXiv new-list pages", "DeepXiv semantic cross-check", "paper HTML/PDF full-text review", "Google and Meta priority reverse search", "official repositories" ], "candidate_pool": { "cross_domain_metadata_total": 363 }, "industrial_recommendation_result": "TAGR passed quantified production A/B; no new Google/Meta paper passed the topic plus full-text online-evidence gate", "candidates": [ { "id": "2608.24034", "key": "tagr", "track": "recommendation", "priority": "P0", "organization": "Kuaishou Technology", "status": "implemented", "doc": "reproductions/2608.24034-tagr/README.md", "evidence_gate": "production A/B: LRE +8.5%, SCC +7.4%, revenue +16.1%" }, { "id": "2608.24053", "key": "wemm-embedding", "track": "foundation-model", "priority": "P0", "organization": "WeChat, Tencent", "status": "implemented", "doc": "reproductions/2608.24053-wemm-embedding/README.md" }, { "id": "2608.24310", "key": "opd-search-plus", "track": "post-training", "priority": "P0", "status": "implemented", "doc": "post-training/2608.24310-opd-search-plus/README.md" }, { "id": "2608.24696", "key": "opdvr", "track": "post-training", "priority": "P0", "status": "implemented", "doc": "post-training/2608.24696-opdvr/README.md" }, { "id": "2608.24870", "key": "spo-plus-plus", "track": "agent", "priority": "P0", "status": "implemented", "doc": "agent-research/2608.24870-spo-plus-plus/README.md" }, { "id": "2608.24747", "key": "skillforge", "track": "agent", "priority": "P0", "status": "implemented", "doc": "agent-research/2608.24747-skillforge/README.md" }, { "id": "2608.24114", "key": "ahead", "track": "agent", "priority": "P0", "status": "implemented", "doc": "agent-research/2608.24114-ahead/README.md" }, { "id": "2608.24571", "key": "smith", "track": "agent", "priority": "P0", "status": "implemented", "doc": "agent-research/2608.24571-smith/README.md" }, { "id": "2608.24295", "track": "recommendation", "priority": "P2", "status": "rejected", "reason": "full text has deployment-oriented analysis but no quantified production A/B or explicit full-traffic rollout" }, { "id": "2608.24132", "track": "recommendation", "priority": "P2", "status": "rejected", "reason": "production migration case study reports fixed-population offline evaluation, not a production online experiment" }, { "id": "2608.24091", "track": "recommendation", "priority": "P2", "status": "rejected", "reason": "no quantified production online evidence" }, { "id": "2608.24079", "track": "recommendation", "priority": "P2", "status": "rejected", "reason": "academic public-data evaluation; no quantified production online evidence" }, { "id": "2608.24176", "track": "recommendation", "priority": "P2", "status": "rejected", "reason": "academic offline evaluation; no quantified production online evidence" } ] }, { "batch": "2026-08-27-daily-increment", "scope_kind": "daily-high-recall-increment-with-one-day-announcement-overlap", "checked_at": "2026-08-28T00:00:00+08:00", "window": "new arXiv listing visible on 2026-08-27; API overlap includes submitted/published 2026-08-26 UTC", "sources": [ "official arXiv API with one-day overlap", "paper HTML/PDF full-text review", "Google and Meta priority reverse search", "official project and repository links" ], "candidate_pool": { "recommendation": 6, "foundation_model": 50, "post_training": 14, "agent": 38 }, "industrial_recommendation_result": "DCEO and TransRetrieval passed quantified production A/B; Meta AMBER was full-text reviewed and rejected for lack of quantified online/full-traffic evidence", "candidates": [ { "id": "2608.25635", "key": "dceo", "track": "recommendation", "priority": "P0", "organization": "Taobao & Tmall Group, Alibaba", "status": "implemented", "doc": "reproductions/2608.25635-dceo/README.md", "evidence_gate": "41-day production A/B: GMV +0.36%" }, { "id": "2608.25528", "key": "transretrieval", "track": "recommendation", "priority": "P0", "organization": "Renmin University of China / Alibaba", "status": "implemented", "doc": "reproductions/2608.25528-transretrieval/README.md", "evidence_gate": "month-long 5% traffic A/B: revenue +2.53%, RPM +1.28%" }, { "id": "2608.26105", "key": "vbvr-pro", "track": "foundation-model", "priority": "P0", "status": "implemented", "doc": "reproductions/2608.26105-vbvr-pro/README.md" }, { "id": "2608.25575", "key": "mllmclip", "track": "foundation-model", "priority": "P1", "status": "implemented", "doc": "reproductions/2608.25575-mllmclip/README.md" }, { "id": "2608.25580", "key": "v-rubrics", "track": "post-training", "priority": "P0", "status": "implemented", "doc": "post-training/2608.25580-v-rubrics/README.md" }, { "id": "2608.25356", "key": "clue-opsd", "track": "post-training", "priority": "P0", "status": "implemented", "doc": "post-training/2608.25356-clue-opsd/README.md" }, { "id": "2608.25243", "key": "grin", "track": "post-training", "priority": "P0", "status": "implemented", "doc": "post-training/2608.25243-grin/README.md" }, { "id": "2608.25583", "key": "grip", "track": "post-training", "priority": "P1", "status": "implemented", "doc": "post-training/2608.25583-grip/README.md" }, { "id": "2608.25593", "key": "jit-agent", "track": "agent", "priority": "P0", "status": "implemented", "doc": "agent-research/2608.25593-jit-agent/README.md" }, { "id": "2608.26086", "key": "traceml", "track": "agent", "priority": "P0", "status": "implemented", "doc": "agent-research/2608.26086-traceml/README.md" }, { "id": "2608.25559", "key": "adavdr", "track": "agent", "priority": "P0", "organization": "Alibaba", "status": "implemented", "doc": "agent-research/2608.25559-adavdr/README.md" }, { "id": "2608.25523", "key": "topas", "track": "agent", "priority": "P0", "status": "implemented", "doc": "agent-research/2608.25523-topas/README.md" }, { "id": "2608.25500", "key": "caskg", "track": "agent", "priority": "P0", "status": "implemented", "doc": "agent-research/2608.25500-caskg/README.md" }, { "id": "2608.25992", "key": "progrouter", "track": "agent", "priority": "P0", "status": "implemented", "doc": "agent-research/2608.25992-progrouter/README.md" }, { "id": "2608.25546", "title": "AMBER", "track": "recommendation", "priority": "P0", "organization": "Meta", "status": "rejected", "reason": "full-text review found no quantified production A/B or explicit full-traffic rollout; organization priority does not waive the evidence gate" }, { "id": "2608.25446", "title": "HSR", "track": "recommendation", "priority": "P2", "status": "rejected", "reason": "academic offline recommendation evaluation; no quantified production online evidence" }, { "id": "2608.25392", "title": "MOTIF", "track": "recommendation", "priority": "P2", "status": "rejected", "reason": "academic offline recommendation evaluation; no quantified production online evidence" } ] }, { "batch": "2026-08-29-cross-domain-incremental", "date": "2026-08-29", "scope_kind": "incremental-full-text-review", "window": "2026-08-27 through 2026-08-29 (one-day announcement overlap)", "required_tracks": [ "recommendation", "foundation-model", "post-training", "agent" ], "priority_institutions": [ "Google", "Meta" ], "institution_review": "No new Google or Meta paper in the increment passed the repository scope and evidence gates.", "source_artifact": "GitHub Actions run 33172409179 plus targeted 2026-08-28 web verification", "candidates": [ { "id": "2608.27413", "key": "friend-gnn", "track": "recommendation", "priority": "P0", "status": "implemented", "doc": "reproductions/2608.27413-friend-gnn/README.md", "evidence_gate": "quantified production A/B: friend additions +16.0%, unique friend adders +11.5%" }, { "id": "2608.27448", "key": "ttpo", "track": "post-training", "priority": "P0", "status": "implemented", "doc": "post-training/2608.27448-ttpo/README.md" }, { "id": "2608.27420", "key": "weak-guide-rlvr", "track": "post-training", "priority": "P0", "status": "implemented", "doc": "post-training/2608.27420-weak-guide-rlvr/README.md" }, { "id": "2608.26735", "key": "uc-mopd", "track": "post-training", "priority": "P0", "status": "implemented", "doc": "post-training/2608.26735-uc-mopd/README.md" }, { "id": "2608.26550", "key": "spear", "track": "post-training", "priority": "P0", "status": "implemented", "doc": "post-training/2608.26550-spear/README.md" }, { "id": "2608.27449", "key": "swe-prime", "track": "agent", "priority": "P0", "status": "implemented", "doc": "agent-research/2608.27449-swe-prime/README.md" }, { "id": "2608.27311", "key": "harnesslens", "track": "agent", "priority": "P0", "status": "implemented", "doc": "agent-research/2608.27311-harnesslens/README.md" }, { "id": "2608.26895", "key": "covemem", "track": "agent", "priority": "P0", "status": "implemented", "doc": "agent-research/2608.26895-covemem/README.md" }, { "id": "2608.26563", "key": "spt", "track": "agent", "priority": "P0", "status": "implemented", "doc": "agent-research/2608.26563-spt/README.md" }, { "id": "2608.27206", "key": "pace-vlm", "title": "PACE", "track": "foundation-model", "priority": "P1", "status": "implemented", "doc": "reproductions/2608.27206-pace-vlm/README.md", "evidence": "real Qwen2.5-VL checkpoint on a public RealWorldQA subset; APC quality/token/latency/memory receipt" }, { "id": "2608.27128", "key": "twinkv", "title": "TwinKV", "track": "foundation-model", "priority": "P1", "status": "implemented", "doc": "reproductions/2608.27128-twinkv/README.md", "evidence": "real Qwen3 KV tensors on public WikiText-2 long contexts; fixed-budget reconstruction/latency/memory receipt" }, { "id": "2608.27006", "title": "Conversational Recommendation over Live E-Commerce Catalogues with Self-Refreshing Retrieval", "track": "recommendation", "priority": "P2", "status": "rejected", "reason": "full-text review found no quantified production A/B or accepted full-traffic evidence" }, { "id": "2608.26579", "title": "Preference Flow Matching with Spectral Factorization for Micro-video Recommendation", "track": "recommendation", "priority": "P2", "status": "rejected", "reason": "offline academic recommendation evaluation without quantified production online evidence" }, { "id": "2608.25381", "title": "MOTIF", "track": "recommendation", "priority": "P2", "status": "rejected", "reason": "offline academic recommendation evaluation without quantified production online evidence" } ] }, { "batch": "2026-08-30-cross-domain-incremental", "date": "2026-08-30", "scope_kind": "incremental-weekend-review", "window": "2026-08-28 through 2026-08-30 (one-day announcement overlap)", "required_tracks": [ "recommendation", "foundation-model", "post-training", "agent" ], "priority_institutions": [ "Google", "Meta" ], "institution_review": "No new in-scope Google or Meta paper was published after the previous waterline.", "source_artifact": "GitHub Actions run 33241816208 plus targeted Google, Google DeepMind, Meta and arXiv weekend verification", "empty_tracks_verified": [ "recommendation", "foundation-model", "post-training", "agent" ], "empty_review": { "source_artifact": "GitHub Actions run 33241816208", "current_manifest_diff": "zero new candidates in all four tracks", "announcement_schedule": "no arXiv weekend announcement on 2026-08-30", "priority_institution_check": "Google, Google DeepMind and Meta public research pages reviewed" }, "candidates": [] }, { "batch": "2026-08-31-late-august-residual-review", "date": "2026-08-31", "scope_kind": "incremental-full-text-residual-review", "window": "2026-08-27 through 2026-08-31 (announcement overlap and residual status review)", "required_tracks": [ "recommendation", "foundation-model", "post-training", "agent" ], "priority_institutions": [ "Google", "Meta" ], "institution_review": "No new Google or Meta paper in the residual batch passed repository scope and evidence gates.", "source_artifact": "GitHub Actions runs 33241816208 and 33297350285 plus full-text review", "candidates": [ { "id": "2608.27455", "key": "criticl", "track": "foundation-model", "priority": "P1", "status": "implemented", "doc": "foundation-models/2608.27455-criticl/README.md" }, { "id": "2608.27409", "key": "rlvr-fusion", "track": "post-training", "priority": "P0", "status": "implemented", "doc": "post-training/2608.27409-rlvr-fusion/README.md" }, { "id": "2608.27065", "key": "video-opsd", "track": "post-training", "priority": "P0", "status": "implemented", "doc": "post-training/2608.27065-video-opsd/README.md" }, { "id": "2608.27032", "key": "normalized-dpo", "track": "post-training", "priority": "P1", "status": "implemented", "doc": "post-training/2608.27032-normalized-dpo/README.md" }, { "id": "2608.27439", "key": "redevoagent", "track": "agent", "priority": "P0", "status": "implemented", "doc": "agent-research/2608.27439-redevoagent/README.md" }, { "id": "2608.27260", "key": "ace-data", "track": "agent", "priority": "P1", "status": "implemented", "doc": "agent-research/2608.27260-ace-data/README.md" }, { "id": "2608.26557", "key": "deeprepro", "track": "agent", "priority": "P0", "status": "implemented", "doc": "agent-research/2608.26557-deeprepro/README.md" }, { "id": "2608.26755", "title": "Residual recommendation candidate", "track": "recommendation", "priority": "P2", "status": "rejected", "reason": "full-text review found no quantified production A/B or explicit full-traffic rollout" }, { "id": "2608.26728", "title": "Meta-review false positive", "track": "recommendation", "priority": "P2", "status": "rejected", "reason": "keyword Meta refers to review terminology rather than Meta company; no industrial online evidence" } ] }, { "batch": "recent-20260905-recommendation", "date": "2026-09-05", "scope_kind": "automated-review-closure", "required_tracks": [ "recommendation" ], "source_artifact": "/tmp/auto-research-discovery-20260905/recommendation-paper-candidates/paper-candidates-recommendation.json", "window": { "start": "2026-08-31", "end": "2026-09-05" }, "candidates": [ { "id": "2609.04047", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "The Dice Roll Method: A Standardized Protocol for Repeated-Query Auditing of Large Language Model Brand Recommendations", "track": "recommendation", "matched_queries": [ "generative-recommendation", "llm-recommendation" ], "source_provenance": [] }, { "id": "2609.03522", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "EPIC: Explicit Posterior Item Conditioning for Semantic ID Diffusion Recommendation", "track": "recommendation", "matched_queries": [ "generative-recommendation", "priority-org-amazon" ], "source_provenance": [] }, { "id": "2609.03369", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "HypRQ-VAE: Hyperbolic Item Indexing for Long-Tail-Aware Generative Recommender Systems", "track": "recommendation", "matched_queries": [ "generative-recommendation", "llm-recommendation", "recsys-general" ], "source_provenance": [] }, { "id": "2609.03313", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SelfDR: Self-Distillation from Reasoning for LLM-Based Recommendation", "track": "recommendation", "matched_queries": [ "generative-recommendation", "llm-recommendation", "recsys-general" ], "source_provenance": [] }, { "id": "2609.03239", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "B2B Customer Conversion Prediction: A Document Representation, Graph Theory, and CatBoost Driven Methodology", "track": "recommendation", "matched_queries": [ "generative-recommendation" ], "source_provenance": [] }, { "id": "2609.02730", "status": "implemented", "priority": "P0", "reason": "纠正摘要级误拒：全文 Section 5.1/Table 1 与 Section 5.2 给出量化线上 A/B、部署配置和容量节省证据；已实现约束反馈闭环。", "key": "coral", "doc": "reproductions/2609.02730-coral/README.md", "full_text_review": { "scope": "full-text", "source_locations": [ "Section 5.1, Table 1", "Section 5.2" ] }, "title": "CORAL: An LLM-Native Harness for Production Recommender Systems", "track": "recommendation", "matched_queries": [ "llm-recommendation" ], "source_provenance": [] }, { "id": "2609.02671", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Recommender System as Slow and Fast Thinkers", "track": "recommendation", "matched_queries": [ "recsys-general" ], "source_provenance": [] }, { "id": "2609.02638", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Oracle, will I ever learn? A study of prediction convergence and complementarity across link prediction models", "track": "recommendation", "matched_queries": [ "recsys-general" ], "source_provenance": [] }, { "id": "2609.02499", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Training seeds and model-selection stability in recommender-system evaluation", "track": "recommendation", "matched_queries": [ "recsys-general" ], "source_provenance": [] }, { "id": "2609.02468", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "DeepAffinity: Long-Term Aspect Preference Prediction in eCommerce using Small Language Models", "track": "recommendation", "matched_queries": [ "generative-recommendation" ], "source_provenance": [] }, { "id": "2609.02170", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "DMRL: Document-Mediated Reinforcement Learning for Skill Optimization in Advertising Recommendation", "track": "recommendation", "matched_queries": [ "recsys-general" ], "source_provenance": [] }, { "id": "2609.02162", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "GenCAR: Generative Counterfactual Alignment with Risk-Controlled Selection for Out-of-Distribution Recommendation", "track": "recommendation", "matched_queries": [ "generative-recommendation" ], "source_provenance": [] }, { "id": "2609.02152", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Beyond Modality Harmony: Orthogonal Purification and Topology-Guided MoE for Conflict-Aware Multimodal Recommendation", "track": "recommendation", "matched_queries": [ "priority-org-amazon", "recsys-general" ], "source_provenance": [] }, { "id": "2609.02062", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SPAR: Enhancing Industrial-Scale Generative POI Recommendation via Real-World Spatial Perception", "track": "recommendation", "matched_queries": [ "generative-recommendation", "llm-recommendation" ], "source_provenance": [] }, { "id": "2609.01761", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Pooling and Drift in Delayed Bandits", "track": "recommendation", "matched_queries": [ "generative-recommendation", "recsys-general" ], "source_provenance": [] }, { "id": "2609.01532", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall", "track": "recommendation", "matched_queries": [], "source_provenance": [ { "source": "Meta Research GitHub", "source_kind": "github-organization", "source_url": "https://api.github.com/orgs/facebookresearch/repos?sort=updated&per_page=100", "organization": "Meta", "relation": "direct", "seed_arxiv_id": null } ] }, { "id": "2609.01273", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Position: Privacy Is a Claim, Not a Property of Synthetic Data", "track": "recommendation", "matched_queries": [ "generative-recommendation" ], "source_provenance": [] }, { "id": "2609.01244", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Post-Training Science for Supervised Fine-Tuning", "track": "recommendation", "matched_queries": [ "generative-recommendation" ], "source_provenance": [] }, { "id": "2609.01240", "key": "rest", "status": "implemented", "priority": "P0", "title": "From Language to Behavior: Scaling Sequence Transformers for Industrial Recommendation Ranking with Rec-Native Designs", "track": "recommendation", "doc": "reproductions/2609.01240-rest/README.md", "matched_queries": [ "advertising-ranking", "deployment-evidence", "industrial-ranking", "llm-recommendation", "production-evidence" ], "source_provenance": [] }, { "id": "2609.01067", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "World Model-Guided Reinforcement Learning via Counterfactual User Engagement Simulation", "track": "recommendation", "matched_queries": [ "llm-recommendation" ], "source_provenance": [] }, { "id": "2609.00986", "key": "tgr", "status": "implemented", "priority": "P0", "title": "TGR: Advancing Industrial Recommendation from Generative-Paradigm Ranking toward Unified Generation and Reasoning", "track": "recommendation", "doc": "reproductions/2609.00986-tgr/README.md", "matched_queries": [ "advertising-ranking", "deployment-evidence", "generative-recommendation", "industrial-ranking", "priority-org-tencent", "production-evidence", "recsys-general" ], "source_provenance": [] }, { "id": "2609.00918", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "RPCBench: A Benchmark for Proactive Premise Critique in LLM-based Recommendation", "track": "recommendation", "matched_queries": [ "llm-recommendation" ], "source_provenance": [] }, { "id": "2609.00913", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SwapRec: Warming Up Cold Items Through Training-Time Swaps", "track": "recommendation", "matched_queries": [ "recsys-general" ], "source_provenance": [] }, { "id": "2609.00618", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Towards Effective Structured Context Modeling for Conversational Recommender Systems via Dual-node Monte Carlo Tree Search", "track": "recommendation", "matched_queries": [ "llm-recommendation", "recsys-general" ], "source_provenance": [] }, { "id": "2609.00434", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SAGE: State-Grounded, Abstention-Aware Evaluation of Task-Oriented Dialogue Agents", "track": "recommendation", "matched_queries": [ "llm-recommendation" ], "source_provenance": [] }, { "id": "2609.00165", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Two-Sided State-Space Models for Sequential Recommendation with Non-Random Multimodal Review Feedback", "track": "recommendation", "matched_queries": [ "generative-recommendation", "priority-org-amazon", "recsys-general" ], "source_provenance": [] }, { "id": "2608.31128", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "DIASENTINEL: An Auditable Multi-Agent System for Guideline-Grounded Diabetes Risk Screening", "track": "recommendation", "matched_queries": [ "llm-recommendation" ], "source_provenance": [] }, { "id": "2608.30873", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Personas Differ from Native-Language Generation: Language Pathways Shape LLM Interpersonal Advice", "track": "recommendation", "matched_queries": [ "llm-recommendation" ], "source_provenance": [] }, { "id": "2609.00086", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Retrieval, Scoring, and Decoding Shape Performance and Stability in LLM-based Conversational Recommendation", "track": "recommendation", "matched_queries": [ "llm-recommendation" ], "source_provenance": [] }, { "id": "2609.01655", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "From Feature Interaction to Feature Transport - A Unified Block for Scalable Recommendation Models", "track": "recommendation", "matched_queries": [ "generative-recommendation" ], "source_provenance": [] }, { "id": "2608.30427", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Ceiling-Clipped Acceptance Histograms Indicate Stranded Speed-up in Block-Diffusion Speculative Decoding", "track": "recommendation", "matched_queries": [ "generative-recommendation" ], "source_provenance": [] }, { "id": "2608.30399", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SemPOI-RL: Aligning LLM Semantic Reasoning for Interpretable Out-of-Town POI Sequential Generation", "track": "recommendation", "matched_queries": [ "llm-recommendation" ], "source_provenance": [] }, { "id": "2608.30333", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Beyond Ranking Accuracy: Evaluating LLM-Cited Feature Rationales for Next Basket Repurchase Recommendation", "track": "recommendation", "matched_queries": [ "generative-recommendation", "llm-recommendation", "recommendation-ranking", "recsys-general" ], "source_provenance": [] }, { "id": "2609.01652", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Not All Matches Are Equally Valuable: An Online Experiment of Retention-Focused Recommendation in a Job-Matching Platform", "track": "recommendation", "matched_queries": [ "recsys-general" ], "source_provenance": [] }, { "id": "2608.30255", "key": "camie", "status": "implemented", "priority": "P0", "title": "CAMIE: Co-Engagement-Aware Multimodal Item Embeddings for Snap Dynamic Product Ads Retrieval", "track": "recommendation", "doc": "reproductions/2608.30255-camie/README.md", "matched_queries": [ "deployment-evidence", "llm-recommendation", "recsys-general" ], "source_provenance": [] }, { "id": "2608.30251", "key": "setmir", "status": "implemented", "priority": "P0", "title": "SetMIR: Multi-Interest Retrieval as Set Prediction", "track": "recommendation", "doc": "reproductions/2608.30251-setmir/README.md", "matched_queries": [ "deployment-evidence", "recsys-general" ], "source_provenance": [] }, { "id": "2608.30186", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "The PUR-1 Cyber-Physical Digital Twin", "track": "recommendation", "matched_queries": [ "recsys-general" ], "source_provenance": [] } ] }, { "batch": "recent-20260905-foundation-model", "date": "2026-09-05", "scope_kind": "automated-review-closure", "required_tracks": [ "foundation-model" ], "source_artifact": "/tmp/auto-research-discovery-20260905/foundation-model-paper-candidates/paper-candidates-foundation-model.json", "window": { "start": "2026-08-31", "end": "2026-09-05" }, "candidates": [ { "id": "2609.04200", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Principia: Relational Physics Tests for Video Models", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.04196", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Puffin-World: Scaling a Unified Multimodal Model with Native 3D World States", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.04180", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Knowledge Acquisition During Pre-training? Large Language Models Learn Better With Auxiliary Views", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2609.04131", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Beyond Retrieval: Progressive Latent Memory Evolution for Streaming Video Understanding", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.04110", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "The Shape of Time: Video-Token Contrast for Temporal Understanding in VideoLMs", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.04096", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Adaptive Vision-Language Grasping via Composable Foundation Priors and Generalizable Grasp Synthesis", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.04083", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CORE: Improving Compositional Reasoning in MLLM Embedding via Reranker Distillation", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.04070", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Continuous Actions from Discrete Minds: Latent-Aligned Planning for End-to-End Autonomous Driving", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.04048", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Translation as a Decision Space: A Multi-Agent Perspective on Low-Resource Dialect Generation", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2609.04034", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Editable Visual Design", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.04013", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "LLM4CKD: Large Language Models for Early Stage Chronic Kidney Disease Screening", "track": "foundation-model", "matched_queries": [ "efficient-inference" ], "source_provenance": [] }, { "id": "2609.04010", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Unlocking Lossless Speedups in LLMs via Discrete Diffusion", "track": "foundation-model", "matched_queries": [ "long-context" ], "source_provenance": [] }, { "id": "2609.03992", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Alignment-Free Text-Audiobox for Voice Dubbing and Full-Duplex Dialogue Synthesis", "track": "foundation-model", "matched_queries": [ "model-compression", "pretraining-data" ], "source_provenance": [] }, { "id": "2609.03985", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "IchthyoNoma: Nomenclature and Context Sensitivity of Zero-Shot Biological Vision--Language Models for Bangladeshi Freshwater Fish Recognition", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.03952", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "WorldReward: Reward Modeling for Camera-Conditioned World Models", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.03949", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "VestigeKV: The NoPE-MLA KV Cache Carries Its Own Eviction Signal in a Vestigial Branch", "track": "foundation-model", "matched_queries": [ "long-context", "model-compression" ], "source_provenance": [] }, { "id": "2609.03941", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Headroom-Drift Replay: A Primitive for Principled Replay Control in GRPO", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2609.03892", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "GraFT: A Training-Free Framework for Spatial Reasoning in Multimodal Large Language Models via 3D Scene Graphs", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.03887", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Beyond Shallow Alignment: How Post-Training Methods Determine Refusal Circuits And Steering Robustness", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.03844", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Flip, Don't Shuffle: Watermarking LLMs at the Speed of Inference", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.03834", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Semantic Bayesian World Models", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.03820", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Select, Compress, Reinvest: A Controlled Study of Visual-Token Allocation in Long-Video MLLMs", "track": "foundation-model", "matched_queries": [ "model-compression", "vision-language" ], "source_provenance": [] }, { "id": "2609.03816", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "When Vision Meets Graphs: A Survey on Graph Reasoning and Learning", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.03811", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "VisCAD: A Foundation Model Suite with Multimodal Industrial CAD Intelligence", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.03806", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SVG-Score: Human-Aligned Evaluation of Text-to-SVG Generation", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.03797", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Transfiver: Human-AI Co-Inference through a Shared Editable State", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.03796", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "LLaDA-Image: Building Strong Image Generators with Fully Open Training Recipes", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.03788", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "A Reverse Sign Language Dictionary: Open-Vocabulary Sign Recognition from Continuous Signing via Video Captioning and Description Retrieval", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.03773", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "RealCADBench: Benchmarking Parametric CAD Modeling from Industrial Design Intents", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.03734", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Beyond BLEU: A Case for Redefining Sign Language Translation Benchmarks", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2609.03729", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Unfold The World: Factorize 4D Properties in Reinforcing Spatial Reasoning", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.03719", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Opening mind by opening architecture: analysis strategies", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.03690", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "MetaStructAtlas: A Grounded 3D Vision-Language Dataset and Benchmark for Functional and Structural Reasoning in Whole-Body PET/CT", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.03677", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Understanding Autonomous Driving Datasets by Describing Differences between Image Subsets in Natural Language", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.03675", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CoFiE: Coarse-to-Fine Evidence Selection for Efficient Streaming Video Understanding", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.03597", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "KhatianDoc: A Human-Verified Benchmark Diagnosing Multimodal LLM Failure on Bengali Legal Land Records", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2609.03595", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "How Far Can Synthetic Data Take Thai OCR?", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.03580", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "HalluPeer: A Taxonomy-driven Benchmark for Detecting Hallucinations in Scientific Peer Reviews", "track": "foundation-model", "matched_queries": [ "long-context" ], "source_provenance": [] }, { "id": "2609.03544", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SafeRI: Recognition and Intervention for Token-Level Safety Intervention in Large Vision Language Models", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.03495", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Spectral characteristics of autoencoder parameters as a vector representation of data", "track": "foundation-model", "matched_queries": [ "llm-architecture", "model-compression" ], "source_provenance": [] }, { "id": "2609.03467", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "When Users Don't Ask: Benchmarking Context-Driven Memory Retrieval in Conversational Agents", "track": "foundation-model", "matched_queries": [ "long-context" ], "source_provenance": [] }, { "id": "2609.03436", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "It's the Problem, Not the Path: Budget and Difficulty Confounds in LLM Reasoning Trajectories", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2609.03430", "status": "implemented", "priority": "P1", "reason": "已实现 prompt 保护、逐 head 随机 KV 淘汰，并完成真实公开 checkpoint/A100 等预算验证。", "key": "random-attention", "doc": "reproductions/2609.03430-random-attention/README.md", "title": "Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2609.03426", "status": "implemented", "priority": "P1", "reason": "已实现离散 n-gram memory、反事实梯度和 GQA zero sink，并完成真实公开 VLM checkpoint/A100 验证。", "key": "lngram-v2", "doc": "reproductions/2609.03426-lngram-v2/README.md", "title": "Lngram v2: Latent N-Gram Memory with Interpretable Discrete Representations", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.03422", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Inferred Generative-Process Diversity Predicts Correlated Failure Across Language Models", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2609.03391", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Exploring the Potential of Contrastive Language-Image Pre-training for Multi-Source Remote Sensing Data", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.03377", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SimpleDesign: A Joint Model for Protein Sequence and Structure Codesign", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.03350", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "From Zero to Hero: An Open LLM Ecosystem for Armenian", "track": "foundation-model", "matched_queries": [ "pretraining-data" ], "source_provenance": [] }, { "id": "2609.03331", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "FPCO-Dialog: A Multi-Turn False-Premise Benchmark for Correction and Cooperation in Vision-Language Models", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.03261", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "MedQA-MM: Shortcuts Behind Medical Visual Reasoning", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.03235", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SGD-KV: Summarization Guided KV Cache Compression", "track": "foundation-model", "matched_queries": [ "efficient-inference", "long-context", "model-compression" ], "source_provenance": [] }, { "id": "2609.03218", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "The Analyst in the Prompt: Role, Retrieval, and Memory Biases in LLM Financial Analysis", "track": "foundation-model", "matched_queries": [ "long-context" ], "source_provenance": [] }, { "id": "2609.03210", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Improving precipitation forecasts in an AI weather model using observational data", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.03206", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Learning to Zoom Efficiently with a Contrastive Curriculum", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.03181", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Jina-OCR-v1: Efficient Document Parsing with Speculative Decoding and Dense Verifiable Rewards", "track": "foundation-model", "matched_queries": [ "model-compression", "vision-language" ], "source_provenance": [] }, { "id": "2609.03177", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Frontier LLMs are effective batch optimizers: Assessing reasoning models in continuous and discrete settings", "track": "foundation-model", "matched_queries": [ "pretraining-data" ], "source_provenance": [] }, { "id": "2609.03160", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "No country for old linguists: LLM-brain alignment underdetermines neural computation", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.03158", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Who Speaks for the Pruned? Visual Token Pruning as Coverage Optimization", "track": "foundation-model", "matched_queries": [ "llm-architecture", "model-compression", "vision-language" ], "source_provenance": [] }, { "id": "2609.03151", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "BASP: Communication-Efficient Batch-Aware Sequence Parallelism for LLM Training", "track": "foundation-model", "matched_queries": [ "long-context" ], "source_provenance": [] }, { "id": "2609.03106", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Scaling Laws, Tabular Data and Actuarial Ratemaking Models", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.03085", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Solving the Needle-in-a-Haystack Problem in Mammography Vision-Language Model with Differentiable Subset Sampling", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.03079", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "LeanStream: A Speculate-and-Refine Streaming Framework for Efficient on-Device LLM Inference", "track": "foundation-model", "matched_queries": [ "efficient-inference" ], "source_provenance": [] }, { "id": "2609.02881", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Graph Machine: Towards Better Pretraining via Edges", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.02780", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "ShallowStream: Index Shallow then Answer Deep for Streaming Video Understanding", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.02772", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "HyperStyler: Low-resource Authorship Style Transfer via Context-aware Style Navigation and Hypernetworks", "track": "foundation-model", "matched_queries": [ "llm-architecture", "model-compression" ], "source_provenance": [] }, { "id": "2609.02749", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.02737", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Language Models Can Control Their Own Attention", "track": "foundation-model", "matched_queries": [ "long-context" ], "source_provenance": [] }, { "id": "2609.02731", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "RVSD: Retrieval Vision Sparse Decoding for Mitigating Visual Hallucinations in Large Vision-Language Models", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.02702", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Trace as State: Reasoning Traces as Conditional States for Long-Context Transformers", "track": "foundation-model", "matched_queries": [ "long-context" ], "source_provenance": [] }, { "id": "2609.02679", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "From Tokens to Semantics: Leveraging Complementary Signals for Hallucination Detection in Black-Box LLMs", "track": "foundation-model", "matched_queries": [ "long-context", "multimodal-llm" ], "source_provenance": [] }, { "id": "2609.02986", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Modern Transformers Are Implicit Hybrids: From Functional Differentiation to Principled Hybrid Architecture Design", "track": "foundation-model", "matched_queries": [ "llm-architecture", "long-context", "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.02664", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Query Rewriting for Complex Object Segmentation in 4D Gaussian Representations", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.02663", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Characterizing Text Branch Sensitivity in Medical Vision-Language Segmentation via Evidence Decoupling", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.02649", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Loom: Weaving Diagnostic Strands into Free-Text Consensus via Embedding-Space Reweighting", "track": "foundation-model", "matched_queries": [ "efficient-inference" ], "source_provenance": [] }, { "id": "2609.02606", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Predictors of Loneliness in Older Adults Using Multimodal Analysis of Speech and Language", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2609.02565", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "MARS: What Retrieval Signals Are Hidden in Multimodal Large Language Models for Text-Video Retrieval?", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.02512", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Beauty is in the AI of the beholder: MLLMs systematically overrate facial attractiveness", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.02496", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Debias-SparseGPT: Bias-Aware Pruning for Large Language Models", "track": "foundation-model", "matched_queries": [ "long-context", "model-compression" ], "source_provenance": [] }, { "id": "2609.02486", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "ViSAR: Training-Free Adaptive-$k$ Retrieval for Visual Document Question Answering", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.02451", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Scalable Kronecker-Fisher Approximation: Efficient Hessian Analysis for Billion-Parameter Language Models Compression", "track": "foundation-model", "matched_queries": [ "llm-architecture", "model-compression" ], "source_provenance": [] }, { "id": "2609.02401", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CA-OPD: Confidence-Aware On-Policy Distillation for Structured Visual Prediction", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.02359", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "TempoGround: State-Aware Streaming Visual Grounding with Vision-Language Models", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.02350", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "LookStep: Efficient Vision-Language Navigation with Linguistic Foresight and Event Driven Memory", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.02341", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Towards Zero-Shot Transfer Across Embodiments For Driving VLAs", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.02339", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "AGI Maze Prediction Datasets: A Compact Benchmark for Learning World Dynamics with Transformers", "track": "foundation-model", "matched_queries": [ "llm-architecture", "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.02318", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "YesTrack: Referring Multi-Object Tracking via MLLM-based Yes/No Verification", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.02293", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SEAL: Reinforcing Global Safety in Mixture-of-Experts through Shared Expert ALignment", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.02264", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Codebook Agent: Amortized Topology Design for LLM Multi-Agent Systems", "track": "foundation-model", "matched_queries": [ "efficient-inference" ], "source_provenance": [] }, { "id": "2609.02255", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "T2LSC-Bench: Benchmarking Localized Semantic Control in Text-to-Image Generation", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.02253", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "APEx: Distillation of Agent Procedural Experience for Adaptive Deep Research Question Answering", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.02251", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Handwriting Trajectory Recovery via Autoregressive Ordered Stroke Instance Prediction", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.02250", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "RideSkill: A Hierarchical Algorithm for Generalized Ride Sharing with LLM-Driven Automatic Evolution", "track": "foundation-model", "matched_queries": [ "efficient-inference" ], "source_provenance": [] }, { "id": "2609.02233", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "InfraPatch: Cross-Task Targeted Grayscale Patch Attacks on Infrared-Adapted Vision-Language Models", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.02207", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "LeakageBench: Document-Level Leakage Risk for Redacting Personally Identifiable Information in Document Images", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.02204", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "TAME: Temporal-Aware Mixture-of-Experts for Text-Video Retrieval", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.02187", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Lightweight Adaptation of General-Purpose VLMs for Multispectral and SAR Image Understanding", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.02163", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Do Cantonese-Adapted Language Models Better Predict Cantonese Reading? A Cross-Model Eye-Tracking Evaluation", "track": "foundation-model", "matched_queries": [ "pretraining-data" ], "source_provenance": [] }, { "id": "2609.02115", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "text2ql: Multi-Target Natural Language Querying via a Language-Agnostic Intermediate Representation", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.02108", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Predict, Don't Iterate: Efficient Adaptive-Length Infilling for Diffusion Language Models", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.02101", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Federated LoRA Adaptation of BiomedCLIP Across Four International Chest X-Ray Cohorts", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "pretraining-data", "vision-language" ], "source_provenance": [] }, { "id": "2609.02088", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Rendering-in-the-Loop: An Execution-Driven Agent for Interactive Web Development", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.02083", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "XMerge: Cross-Axis Selection and Reconstructive Layer Merging for LLM Depth Compression", "track": "foundation-model", "matched_queries": [ "llm-architecture", "model-compression", "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.02082", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Transfer Safety Awareness for Cross-Modal Safety Drift in Multimodal Large Language Models", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2609.02059", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "DocHop: Benchmarking Out-of-domain Multi-hop Reasoning in Information-Dense Documents", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.02049", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "The Dynamics of Continuous Mixture Collapse in Language Models", "track": "foundation-model", "matched_queries": [ "llm-architecture", "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.02039", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Test-Time Logit Prompting for Source-Free Missing Modality Adaptation", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.02028", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Detecting Object Hallucinations in Large Vision-Language Models via Cross-Modal Attention Drifts and Mask-Based Verification", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.02000", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Who Drives the Probability Game of VLMs? A Temporal Causal Drive Evaluation Framework", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.01971", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "NS-Copilot: An LLM-Driven Agent System for Autonomous Neuroscience Analysis", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.01962", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Post-Training Ternarization of Qwen3-4B Capability, Effective Bit Budget, Storage Compression, and Deployment", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2609.01933", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "OR-Transformer: Scaling Real-Time Decision-Making to 1,000 Items", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.01925", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CRISP: Cliff-awaRe Input-adaptive Sparse Prefilling with Structural-Mass-Motivated Routing", "track": "foundation-model", "matched_queries": [ "long-context" ], "source_provenance": [] }, { "id": "2609.01918", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Grounded, Compute-Efficient LLM Policy Agents for Energy-Poverty Equity in Physically-Constrained Peer-to-Peer Energy Markets", "track": "foundation-model", "matched_queries": [ "efficient-inference", "model-compression" ], "source_provenance": [] }, { "id": "2609.01888", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Does Playing it Safe Count as Faithfulness? Reassessing LVLM Hallucination Mitigation Methods", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.01871", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Latent unified smooth Hamiltonians for excited state chemistry", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.02947", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Privacy-Preserving Heterogeneous Multi-LLM Federated Inference for Cognitive Diagnosis", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.01816", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Video2Reaction: Training Foundation Video Models to Predict Audience Reaction", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.01807", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "hLLM: Single Pass Decoding for Generative Reranking", "track": "foundation-model", "matched_queries": [ "efficient-inference", "llm-architecture" ], "source_provenance": [] }, { "id": "2609.01798", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "How Do Prompt Variations Affect Energy Consumption in On-Device LLMs?", "track": "foundation-model", "matched_queries": [ "efficient-inference" ], "source_provenance": [] }, { "id": "2609.01795", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "DESA-TTA: Dynamic EMA and Source Anchoring for Test-Time Adaptation", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.01786", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Ten Architectures, One Error: Shared Failure Modes in Hyperspectral Classification under Spatially Disjoint Evaluation", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.01772", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "MemeCULT-1K: Benchmarking South Asian Cultural Context and Humor Understanding of Multimodal Models", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.01768", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Emergence of Fibrations, Compression, and Symmetry Breaking in Artificial Neural Networks", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.01757", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "AlphaRAD: Grounded Zero-Shot Classification in Chest Radiology via $α$-Corrected Binary Cross Entropy and Factorized Latent Supervision", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.01607", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Uncovering Understanding-Generation Synergy in Native Unified Multimodal Models: From Representation, Task to System", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.01604", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Beyond Scores: Understanding LLM-as-a-Judge Mechanisms in Summarization Evaluation", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.01587", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "The Structure of Quantization Damage in LLMs: Why the Next Bit Should Be Spent Globally", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.01582", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SpatialGuard: Harness-Guided Verifiable Spatial Reasoning for Text-to-Image Generation", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.01567", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Selective Agent Guidance via Entropy: Learning Autonomous Policies from Imperfect VLM Teachers", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.01560", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "H3-World: Turning Language Understanding into World Control", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.01532", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall", "track": "foundation-model", "matched_queries": [], "source_provenance": [ { "source": "Meta Research GitHub", "source_kind": "github-organization", "source_url": "https://api.github.com/orgs/facebookresearch/repos?sort=updated&per_page=100", "organization": "Meta", "relation": "direct", "seed_arxiv_id": null } ] }, { "id": "2609.01515", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "TempCloze: Can Video-LLMs Identify the Missing Middle?", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.01507", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "LatentPress: Context Compression Beyond Text and Vision", "track": "foundation-model", "matched_queries": [ "long-context", "model-compression" ], "source_provenance": [] }, { "id": "2609.01691", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "FairLens: Benchmarking Fairness in Vision-Language Models for High-Stakes Decision-Making", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.01470", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "RadMatch: Auditable Radiology Report Evaluation via Finding-Level Matching", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.01417", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Predicting Subsurface Abnormalities Growth using Physics-Informed Neural Networks", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.01409", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "EdiTikZ: Scientific Figure Editing from Revision Trajectories", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.01383", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "InSight: A Benchmark for Agentic Claim Verification in Interactive Visualizations", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.01379", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Polish ModernBERT: The Long and Short of Polish Language Understanding", "track": "foundation-model", "matched_queries": [ "llm-architecture", "long-context", "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.01375", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "IntroConformal: Conformal Factuality Guarantees for Large Vision-Language Models via Introspective Signals", "track": "foundation-model", "matched_queries": [ "llm-architecture", "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.01343", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.01341", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Probing Factual Knowledge Transfer with Training Data Interventions", "track": "foundation-model", "matched_queries": [ "pretraining-data" ], "source_provenance": [] }, { "id": "2609.01325", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "VerTox: Verifiable Reward-Guided Corpus Poisoning Against Neural Ranking Models", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.01318", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Reliability Challenges in Diffusion Vision-Language Models", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.01289", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Agentic Multimodal Models for Environmental Hyperspectral Unmixing", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.01262", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Solving In-Table Prediction Problems by Deep Neural Networks with Performance Evaluation Using Synthetic Data", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.01231", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Multi-Head Self Attention is a Parameter Identification Mechanism", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.01224", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "S$^2$Prune: Spatially Structured Visual Token Pruning for Multimodal Large Language Models", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.01212", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Recent Developments in Transformer Inference Deployment on FPGA Platforms: A Survey", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.01200", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Compressing AI Traffic: Standardized Neural Network Coding of Visual-Token Representations in Split Vision-Language Inference", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.01170", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Pre-carved Niches: The Formation Dynamics of Modular Task Partitions in Early LLM Training", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.01147", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "On the Design Fundamentals of Pixel Text Representation Learning", "track": "foundation-model", "matched_queries": [ "model-compression", "vision-language" ], "source_provenance": [] }, { "id": "2609.01136", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Different Changes Require Different Reasoning: Change-Type-Specialized Experts for Robust Change Captioning", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.01059", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Dyn-3D: Unveiling and Resolving Ego-Motion Ambiguity in Vision-Language Models", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.01030", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Web Price Extraction: State of the Art and an Adaptive Browserless Implementation", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.01024", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "PCoMoE: Shifting MoE Inference from Monolithic Expert Selection to Fine-Grained Path Composition", "track": "foundation-model", "matched_queries": [ "efficient-inference", "llm-architecture" ], "source_provenance": [] }, { "id": "2609.01004", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SinkPruner: Sink-Free Visual Token Pruning for Multimodal Large Language Models", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.00948", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "From Terminology to Diagrams: Visual-Instruction Generation for Scientific Diagram Understanding", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.00918", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "RPCBench: A Benchmark for Proactive Premise Critique in LLM-based Recommendation", "track": "foundation-model", "matched_queries": [ "long-context" ], "source_provenance": [] }, { "id": "2609.00909", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "A multicenter benchmark and clinically structured metric for coronary CTA report generation", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00898", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Vision-Language-Guided Pseudo-Labels for Unsupervised Domain Adaptation in Semantic Segmentation for Waste Sorting", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00885", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Denoising Diffusion Generative Models Secretly Calculate Attentions", "track": "foundation-model", "matched_queries": [ "llm-architecture", "transformer-architecture", "vision-language" ], "source_provenance": [] }, { "id": "2609.00868", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "The Visual Insensitivity Gap: Diagnosing When Vision-Language Models Fail to Use Visual Evidence", "track": "foundation-model", "matched_queries": [ "llm-architecture", "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.00866", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Benchmarking Vision-Language Models for Automated Pathology Diagnosis and Report Generation", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.00830", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Visual Attention Faithfulness in Vision-Language Models is Heterogeneous", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00816", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Can Scene Text Recognition Read Rare Compositions?", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00802", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "TEIDAN: A Multilingual Multiparty Dialogue Corpus", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2609.00796", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SFAD: Speculative Factuality-Aware Decoding", "track": "foundation-model", "matched_queries": [ "efficient-inference" ], "source_provenance": [] }, { "id": "2609.00791", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Instella-MoE Technical Report", "track": "foundation-model", "matched_queries": [ "llm-architecture", "long-context" ], "source_provenance": [] }, { "id": "2609.00788", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Forbid Your Attention: Fooling Multimodal Large Language Models by Selectively Removing Intrinsic Focus in Spectral Domain", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.00776", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Solaris: Towards Interfaces That Are Generated, Not Coded", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00764", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Are You Thinking What I am Thinking? : Examining Conceptual Separation in Neural Architectures", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.00759", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Compile, Don't Memorize: A Context Compilation Architecture (CCA) for In-Context Learning", "track": "foundation-model", "matched_queries": [ "llm-architecture", "long-context" ], "source_provenance": [] }, { "id": "2609.00756", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Joint Training Is Not Enough: Conditioned Cross-Granularity Training for Multimodal Document Understanding", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2609.00747", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Can Large Language Models Forecast What Researchers Study Next?", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2609.00746", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Text Capability Loss in Vision-Language Adaptation: An Attention-Sink Diagnosis", "track": "foundation-model", "matched_queries": [ "pretraining-data" ], "source_provenance": [] }, { "id": "2609.00712", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "EarthLD: Towards Unified Open-World Landslide Understanding via Vision-Language Guided Diffusion Models", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00709", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Controllable Image Captioning with Prompt-Conditioned Scene Rewards", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00667", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "From Saliency to Discriminability: Rank-Preserving Visual Token Pruning for VLM Rerankers", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00663", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Separating perception from reasoning in vision-language models: a model-free render ceiling for crystal structures", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.00661", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Do Satellites See Commuters? A Critical Benchmark of Vision Foundation Models", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00658", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Teaching Vision-Language Models to Use the Scale They Are Given: Label-Free Equivariance Training for Metric Physical Reasoning", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00654", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SciTrue: Reliable Scientific Claim Validation with Frontier and Open Language Models at the NTCIR SciClaimEval Task", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2609.00649", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "You Cannot Photograph the Same Street Twice: Reliability Limits in Vision-Language Measurement of Urban Change", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00629", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "ExpArt-KG: Artwork Image Description Generation through Iterative Exploration of Knowledge Graphs", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00628", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Restrict, Don't Retrain: Inference-Time VLM Guidance for Zero-Shot Aerial Segmentation", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00615", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "BeamRMX: Radiation-Pattern-Driven Learning for Generalizable Beam Radio Map Prediction and Beam Management", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.00605", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Confess What You Know: Forget-Set Misalignment with Model Knowledge in LLM Unlearning", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2609.00593", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "BrainDiff: Longitudinal Report Generation for Multimodal Brain MRI", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.00591", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "A Glance Is All You Need: Single-Pass Fine-Grained Image Captioning with SimLoss", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00590", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CRAFT: Fine-Tuning Pre-hoc Explainability in AI-native 6G RAN", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.00581", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Enoki: Efficient Multi-Level Hallucination Detection", "track": "foundation-model", "matched_queries": [ "efficient-inference" ], "source_provenance": [] }, { "id": "2609.00570", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "VoiceLongMemEval: Do Assistants Remember How You Sounded?", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.00555", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Potential-Guided Particle Steering for Negation-Constrained Dexterous Grasping", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00551", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "EM^2Mem: Event-Centric Multimodal Memory for Large Language Models", "track": "foundation-model", "matched_queries": [ "long-context", "multimodal-llm" ], "source_provenance": [] }, { "id": "2609.00550", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Same Semantics, Different Outcome: On the Modality Robustness of Multimodal LLMs under Knowledge Conflict", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2609.00505", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "ViTAL-X: Video-Text Alignment with Cross-Modal Temporal Edits", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00476", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Less Is More: Balancing Positive and Negative Space in Visual Concept Blending", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.00474", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Exploring Collaboration between a language and a non-language agent", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2609.00470", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "TRIS: A Tri-Layer Retrieval Integrity Sieve Against Knowledge Poisoning", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.00454", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Location-Aware Language Models via Secondary Embeddings", "track": "foundation-model", "matched_queries": [ "llm-architecture", "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.00450", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "HBQ: Hierarchical Scaling Block Quantization with Hardware-Efficiency-Aware Design for Accurate LLM Inference", "track": "foundation-model", "matched_queries": [ "efficient-inference", "llm-architecture" ], "source_provenance": [] }, { "id": "2609.00447", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Instance-Guided Report Anchoring for Text-Free 3D Abnormality Segmentation in Chest CT", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00411", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Unmasking Face Embeddings: Reading, Rendering and Naming with Foundation Models", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00407", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "DynaNDE: Dynamic Near-Data Expert Scheduling for Batched MoE Inference", "track": "foundation-model", "matched_queries": [ "efficient-inference", "llm-architecture" ], "source_provenance": [] }, { "id": "2609.00403", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "A Multi-Branch Feature Fusion Approach for Health Misinformation Detection and Propagation", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.00396", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SlideMix: Enhancing Whole Slide Image Analysis via Multimodal Shuffling", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.00377", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "FoldingAgent: Inferring Parametric Origami Procedures from Demonstration Videos", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00374", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Adapting Without Gradients: Affine Statistics Transport and What Its Certificate Can Tell You", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.00355", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Vision Is Not Overhead: One-Pass Block Drafting for Lossless Speculative Decoding in Vision-Language Models", "track": "foundation-model", "matched_queries": [ "model-compression", "vision-language" ], "source_provenance": [] }, { "id": "2609.00297", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Geometry-aware Latent Autoregressive Generative Model for PDEs in Complex Domains", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.00293", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Slow to See, Slow to Suppress: Understanding the Effects of Modality in Context-Memory Conflicts", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.00284", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "WiSDoM: Wireless Sparse Decision Transformer with Mixture-of-Experts for Multi-Task Mobile Network Optimization", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.00242", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CoLT-Drive: Counterfactual Long-Tail Benchmarking and Knowledge-Preserving Adaptation for Driving Affordance Prediction", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00237", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Learning What to Retain: Gated-Memory Routing for Efficient Collaboration in Multi-Agent LLM Systems", "track": "foundation-model", "matched_queries": [ "efficient-inference" ], "source_provenance": [] }, { "id": "2609.00232", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Beyond Blind Compliance: Benchmarking Task Verification in OCR Reasoning", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.00231", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Beyond Language Priors: Diagnosing and Fixing Visual-Origin Hallucinations in Multimodal LLM", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.00228", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Bridging Lexical Divergence: LLM-Assisted, Cost-Efficient, Zero-shot Scientific Entity Linking", "track": "foundation-model", "matched_queries": [ "pretraining-data" ], "source_provenance": [] }, { "id": "2609.00224", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "QTEA: Ternary LLMs with Sparse Residual Salient Weight and By-Column Optimization", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2609.00206", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Distributed Implicit Harm: A Compositional Safety Blind Spot in MLLM-Based Video Moderation", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.01659", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Beyond Textual Chain-of-Thought: A Survey on Action-Grounded Reasoning in Autonomous Driving", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00192", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "LLM-Driven Autonomous Vehicles Inherit Human Driver Biases in Pedestrian Yielding: Results and Implications From A New Benchmark", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00184", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Synthetic Worlds for Temporal Evaluation and Knowledge Updating in LLMs", "track": "foundation-model", "matched_queries": [ "pretraining-data" ], "source_provenance": [] }, { "id": "2609.00111", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Qwen-Drive-1.0: An Initial Step towards a Vision-Language Foundation Model for Autonomous Driving", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2608.31119", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "PaperGym: Rubric-Centered Evolution for Research-Plan Generation", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2608.31100", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement?", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2609.00103", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Good Memory Has ECC: Evaluating the Memory of Vision-Language Models Beyond Accuracy", "track": "foundation-model", "matched_queries": [ "llm-architecture", "model-compression", "transformer-architecture" ], "source_provenance": [] }, { "id": "2608.31066", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Every Token Leaves a Ripple in the Stream of Thought: Eliciting Model-Internal Token Saliency for Chain-of-Thought Compression", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2608.31052", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Segmentation of Bovid Dentition Under Imperfect Annotations: A Comparative Study of Convolutional and Attention Models", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2608.31045", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Rotational Equivariance in Machine Learning: A Comprehensive Tutorial", "track": "foundation-model", "matched_queries": [ "llm-architecture", "transformer-architecture" ], "source_provenance": [] }, { "id": "2608.31014", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Evidence-Bounded Mental Health Reasoning from Heterogeneous Speech Protocols", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2608.30997", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Multi-View Reflective Surface Inspection via Semantic-Saliency Cross-Verification", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2608.30996", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Faithfulness Is Not Free: Auditing Offline KV-Cache Quantization in Retrieval-Augmented Generation", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2608.30986", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Controlling Refusal Behavior of LLMs via Stiefel-Constrained Rotation Steering", "track": "foundation-model", "matched_queries": [ "efficient-inference" ], "source_provenance": [] }, { "id": "2608.30964", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Vision Models Predict Urban Scene Appraisal with Limited Neural Alignment", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2608.30963", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "A Universal Context-Reuse Layer for Cross-Model KV Sharing", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2608.30959", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "LOCI: A Locator-Critic with Refinement Loop", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.00099", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Generative artificial intelligence for reliable mechanistic reasoning for corrosion", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.00097", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Faster Than Flash: Exploiting Attention Sparsity for Efficient Long-Context Decoding", "track": "foundation-model", "matched_queries": [ "long-context" ], "source_provenance": [] }, { "id": "2608.30912", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Responsible Integration of AI in Cancer Genomics: Barriers, Risks, and Pathways to Trustworthy Clinical Translation", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2608.30903", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "MMDS-Bench: Benchmarking Multimodal Large Language Models on Dynamic Stance in Social Media Interactions", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2608.30902", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Low-Resource Preference Adaptation of LLMs via Activation-Based Label Propagation", "track": "foundation-model", "matched_queries": [ "pretraining-data" ], "source_provenance": [] }, { "id": "2608.30884", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Evaluating and Mitigating Anti-LGBTQ Biases in German and Multilingual Language Models", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2608.30857", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "TAMI: Temporally Aligned, Missingness-Aware, and Interpretable Multimodal Fusion for Mental Health Assessment in Older Adults with Mild Cognitive Impairment", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2608.30853", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Linguistic Distance Segregates Latent Representations in Automatic Speech Recognition Systems", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2608.30841", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "HSRM: Hidden-State Reward Models for Test-Time Verification", "track": "foundation-model", "matched_queries": [ "efficient-inference" ], "source_provenance": [] }, { "id": "2608.30811", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "TopoCompress: Long Context Compression via Graph-Wired Semantic Trajectories", "track": "foundation-model", "matched_queries": [ "long-context", "model-compression" ], "source_provenance": [] }, { "id": "2608.30804", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Geometric Attractor Monitoring: A Robust and Frugal Framework for Multi-modal Industrial Robotic Cycles", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2608.30758", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CheXGround: Anatomical Region Tokens for Grounded Longitudinal Chest X-ray Interpretation", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2608.30751", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Autoregressive Mosaics: Probing 2D Spatial Reasoning in Text-Only Language Models", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2608.30750", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Do VLMs Share Safety Neurons Across Modalities?", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.01657", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "NeoMME: A Single-Tower Multimodal-Native Multilingual Foundation Encoder for Efficient Fine-Tuning and Inference", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2608.30725", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Where Do Multilingual Vision-Language Encoders Fail on Low-Resource Languages?", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2608.30716", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SocialReasonBench: A Video-QA Benchmark for Social Reasoning with Counterfactual Narrative Videos", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2608.30712", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "GUIDE: Guiding Internal Evidence with Language Instructions", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2608.30705", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "VisLens: Single-Pass Interpretable Visual Search for Multimodal LLMs", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2608.30703", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SingProbe Technical Report", "track": "foundation-model", "matched_queries": [ "efficient-inference" ], "source_provenance": [] }, { "id": "2608.30678", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "OCR-MetaReasoning Benchmark: Evaluating the Meta-Reasoning Ability of MLLMs in Text-Rich Image Understanding", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2608.30653", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Fine-Grained Multi Image Object Hallucination Benchmark", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2608.30649", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Where Identity Lives: Localized, Retain-Free Identity Unlearning in Multimodal Large Language Models", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2608.30647", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "What It Costs to Compose, Rebuild, and Correct Precomputed Memory", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2608.30646", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "BiG-SURE - Bipartite Graph for Semantic Uncertainty and Reliability Estimation of LLMs", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2608.30627", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "REER-PT: Reverse-Engineered Reasoning for Perplexity-Guided Pre-training Data Augmentation", "track": "foundation-model", "matched_queries": [ "llm-architecture", "transformer-architecture" ], "source_provenance": [] }, { "id": "2608.30621", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Cost-efficient Active Learning for Referring Image Segmentation and Grounding", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2608.30550", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "GarmentWeaver: Schema-Aware Structured Synthesis for Multimodal Sewing Patterns", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2608.30510", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Lot Machine: Multimodal Lot Extraction from Auction Catalogs", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2608.30505", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Tensor Methods for Language Models: From Token Representation to Training, Adaptation, Inference, Compression, and Interpretability", "track": "foundation-model", "matched_queries": [ "efficient-inference", "model-compression" ], "source_provenance": [] }, { "id": "2608.30480", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "VisER: Visual Evidence and Reliance for Object Hallucination Detection in LVLMs", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2608.30478", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Agents in the Large: Perception-Centered Architecture for Persistent Agents", "track": "foundation-model", "matched_queries": [ "llm-architecture", "long-context" ], "source_provenance": [] }, { "id": "2608.30475", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "ImageEval 2026: Culturally Grounded Arabic Multimodal Evaluation", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2608.30462", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Enhancing Low-Resource Language Reasoning via High-Resource Language Feature Transfer", "track": "foundation-model", "matched_queries": [ "pretraining-data" ], "source_provenance": [] }, { "id": "2608.30461", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "From Final Artifacts to Trajectories: Retrospective Process Supervision for Evidence-Grounded Long-Form Generation", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2608.30439", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Event-Driven Language Models with Sparse Neural Activity for Neuromorphic Hardware", "track": "foundation-model", "matched_queries": [ "llm-architecture", "transformer-architecture" ], "source_provenance": [] }, { "id": "2608.30429", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "EvoSkill Injection: Red-Teaming Autonomous Skill Generation and Evolution in Self-Evolving Agents", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2608.30428", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Lies We Can See: Joint Verbal and Non-Verbal Deception by VLM Agents in Embodied Social Interactions", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2608.30420", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Whole-Slide Image Analysis under Realistic Few-Shot Annotation Protocols", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2608.30398", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Beyond Polarization: The Generative Constraint of Chain-of-Thought in Pointwise Reranking", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2608.30388", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "PRISM: Predictive Recomposition via Semantic Latent Decomposition for View-invariant Video Representation Learning", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2608.30320", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "On the Design of Qwen3.8-Next Architecture: Evaluation, Efficiency, and Training Stability", "track": "foundation-model", "matched_queries": [ "llm-architecture", "model-compression" ], "source_provenance": [] }, { "id": "2608.30315", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Context Staircase: Signature-Aligned Dynamics of Token Embeddings under Small Initialization", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2608.30310", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Tail-Replay: Escaping the Curse of Linear Attention in Prefix Caching for Hybrid LLMs", "track": "foundation-model", "matched_queries": [ "efficient-inference", "long-context", "model-compression" ], "source_provenance": [] }, { "id": "2608.30307", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "ScenePilot: Grow-and-Repair Policy for Text-Driven 3D Indoor Scene Generation", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2608.30295", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CateKV: On Sequential Consistency for Long-Context LLM Inference Acceleration", "track": "foundation-model", "matched_queries": [ "long-context" ], "source_provenance": [] }, { "id": "2608.30270", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Read the Room, Read the Image: Understanding Indirect Speech Acts in Multimodal Visual Contexts", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2608.30263", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Centering before Pruning: Lightweight Geometry Correction for Diversity-Based Visual Token Pruning in LVLMs", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2608.30260", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Using Prosody to Predict Syntactic Structure", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2608.30226", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "LaMoC: Loss-Aware Modular Compression for LLMs", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2608.30213", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Towards a Joint Khmer Text Recognition and Word Segmentation", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2608.30210", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Frontier vision-language models have overtaken young adults at detecting AI-generated portraits -- but not their calibration", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2608.30209", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "DICS: Exploring Data Intrinsic Consistency for Visual Instruction Selection", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2608.30204", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "When Models Hear What They Expect: Diagnosing Prosodic Heuristics in Multimodal Sarcasm Detection", "track": "foundation-model", "matched_queries": [ "llm-architecture", "multimodal-llm" ], "source_provenance": [] }, { "id": "2608.30181", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "A.X K2 Technical Report", "track": "foundation-model", "matched_queries": [ "long-context" ], "source_provenance": [] }, { "id": "2608.30156", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Reactivating Test-Time Scaling for Plane Geometry Problem Solving", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2608.30135", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Verification-Aware Training for Speculative Decoding", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2608.30125", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "VIBE: Video Instruction-aligned Background music gEneration", "track": "foundation-model", "matched_queries": [ "llm-architecture", "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2608.30110", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Can LLMs Take the Pulse of the Economy? A Real-Time Evaluation of LLM Nowcasts on Macroeconomic Indicators", "track": "foundation-model", "matched_queries": [ "pretraining-data" ], "source_provenance": [] }, { "id": "2608.30105", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "A Simple Transformer Pipeline for Full-Key Side-Channel Attacks on Uncropped Datasets", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] } ] }, { "batch": "recent-20260905-post-training", "date": "2026-09-05", "scope_kind": "automated-review-closure", "required_tracks": [ "post-training" ], "source_artifact": "/tmp/auto-research-discovery-20260905/post-training-paper-candidates/paper-candidates-post-training.json", "window": { "start": "2026-08-31", "end": "2026-09-05" }, "candidates": [ { "id": "2609.04194", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Legibility is Not Interpretability: Comparing Judged and Actual Importance in Chain-Of-Thought Reasoning", "track": "post-training", "matched_queries": [ "process-reward", "reward-model" ], "source_provenance": [] }, { "id": "2609.04173", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Last Translation Benchmark", "track": "post-training", "matched_queries": [ "reward-model" ], "source_provenance": [] }, { "id": "2609.04172", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Rethinking On-Policy Distillation of Large Language Models II: One Training Example", "track": "post-training", "matched_queries": [ "on-policy-distillation", "post-training" ], "source_provenance": [] }, { "id": "2609.04061", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "When Models Edit Too Much: On the Fidelity of Minimal Code Edits", "track": "post-training", "matched_queries": [ "llm-rl", "post-training" ], "source_provenance": [] }, { "id": "2609.03960", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "FiMI Banking: A Sovereign Model for Indian Retail Banking", "track": "post-training", "matched_queries": [ "llm-rl", "post-training", "preference-optimization", "reward-model" ], "source_provenance": [] }, { "id": "2609.03955", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Two-Stage Reinforcement Learning for Sound and Adversarial Test Generation in Code LLMs", "track": "post-training", "matched_queries": [ "llm-rl", "test-time-rl" ], "source_provenance": [] }, { "id": "2609.03941", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Headroom-Drift Replay: A Primitive for Principled Replay Control in GRPO", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.03894", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CROCODIL: Cross-Model Code Editing with LLMs", "track": "post-training", "matched_queries": [ "post-training", "reward-model" ], "source_provenance": [] }, { "id": "2609.03887", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Beyond Shallow Alignment: How Post-Training Methods Determine Refusal Circuits And Steering Robustness", "track": "post-training", "matched_queries": [ "post-training", "preference-optimization" ], "source_provenance": [] }, { "id": "2609.03846", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "EF1-Constrained Nash Social Welfare with Identical Additive Valuations: Complexity, Guarantees, and Experiments", "track": "post-training", "matched_queries": [ "test-time-rl" ], "source_provenance": [] }, { "id": "2609.03811", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "VisCAD: A Foundation Model Suite with Multimodal Industrial CAD Intelligence", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.03342", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Gradients Know What Outcomes Don't: Unlocking Reinforcement Learning for LLM Reasoning with Gradient-Aligned Rewards", "track": "post-training", "matched_queries": [ "llm-rl", "process-reward", "reward-model" ], "source_provenance": [] }, { "id": "2609.03324", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "DE-Venus: A Data-Efficient RLVR Framework for Large Language Models", "track": "post-training", "matched_queries": [ "llm-rl", "reward-model" ], "source_provenance": [] }, { "id": "2609.03215", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SWIM: Student Writing Simulation via Proficiency-Conditioned Generation", "track": "post-training", "matched_queries": [ "llm-rl", "reward-model" ], "source_provenance": [] }, { "id": "2609.03206", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Learning to Zoom Efficiently with a Contrastive Curriculum", "track": "post-training", "matched_queries": [ "reward-model" ], "source_provenance": [] }, { "id": "2609.03181", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Jina-OCR-v1: Efficient Document Parsing with Speculative Decoding and Dense Verifiable Rewards", "track": "post-training", "matched_queries": [ "post-training", "reward-model" ], "source_provenance": [] }, { "id": "2609.02849", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Post-Training Language Models for Gold-Medal Performance in Coding Competitions", "track": "post-training", "matched_queries": [ "llm-rl", "post-training", "test-time-rl" ], "source_provenance": [] }, { "id": "2609.02817", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Cliff: Learning Process Rewards from the First Mistake", "track": "post-training", "matched_queries": [ "llm-rl", "on-policy-distillation", "post-training", "process-reward", "reward-model" ], "source_provenance": [] }, { "id": "2609.02677", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Eliciting ESG Preferences for Reinforcement Learning-Based Portfolio Optimization", "track": "post-training", "matched_queries": [ "llm-rl", "preference-optimization" ], "source_provenance": [] }, { "id": "2609.02548", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Learn from Whoever Is Right: Answer-Verified Multi-Teacher Distillation for Multi-Domain LLMs", "track": "post-training", "matched_queries": [ "llm-rl", "on-policy-distillation" ], "source_provenance": [] }, { "id": "2609.02496", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Debias-SparseGPT: Bias-Aware Pruning for Large Language Models", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.02366", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "NE-R1: Enhancing Named Entity Recognition Model via Reinforcement Learning", "track": "post-training", "matched_queries": [ "llm-rl", "reward-model" ], "source_provenance": [] }, { "id": "2609.02293", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SEAL: Reinforcing Global Safety in Mixture-of-Experts through Shared Expert ALignment", "track": "post-training", "matched_queries": [ "llm-rl" ], "source_provenance": [] }, { "id": "2609.02275", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Do Large Language Models Capture the Diversity in their Training Data?", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.02273", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CoMerge: Conflict-Driven Preference Optimization for Multi-Task Model Merging", "track": "post-training", "matched_queries": [ "preference-optimization" ], "source_provenance": [] }, { "id": "2609.02262", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "From Detection to Characterization: A Large-Scale Study of Ragebait on Japanese X", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.02253", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "APEx: Distillation of Agent Procedural Experience for Adaptive Deep Research Question Answering", "track": "post-training", "matched_queries": [ "llm-rl", "on-policy-distillation", "opd", "reward-model", "test-time-rl" ], "source_provenance": [] }, { "id": "2609.02250", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "RideSkill: A Hierarchical Algorithm for Generalized Ride Sharing with LLM-Driven Automatic Evolution", "track": "post-training", "matched_queries": [ "llm-rl" ], "source_provenance": [] }, { "id": "2609.02170", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "DMRL: Document-Mediated Reinforcement Learning for Skill Optimization in Advertising Recommendation", "track": "post-training", "matched_queries": [ "llm-rl", "post-training", "process-reward", "reward-model" ], "source_provenance": [] }, { "id": "2609.02089", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "IDEEA: training-free Input-Dependent stEEring via Activation cluster matching", "track": "post-training", "matched_queries": [ "llm-rl" ], "source_provenance": [] }, { "id": "2609.02083", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "XMerge: Cross-Axis Selection and Reconstructive Layer Merging for LLM Depth Compression", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.02042", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Act More, Decide Less: Skill-Guided Adaptive Action Chunking for Long-Horizon LLM Agents", "track": "post-training", "matched_queries": [ "llm-rl", "on-policy-distillation" ], "source_provenance": [] }, { "id": "2609.01962", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Post-Training Ternarization of Qwen3-4B Capability, Effective Bit Budget, Storage Compression, and Deployment", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.01867", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Thinking effort aligns between humans and reasoning models in abductive reasoning", "track": "post-training", "matched_queries": [ "llm-rl", "preference-optimization", "reward-model", "test-time-rl" ], "source_provenance": [] }, { "id": "2609.01597", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "The Rise of Verbal Reinforcement Learning", "track": "post-training", "matched_queries": [ "llm-rl", "reward-model", "test-time-rl" ], "source_provenance": [] }, { "id": "2609.01596", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Facet-0: A Robotic Foundation Model for Contact-Rich Precise Manipulation", "track": "post-training", "matched_queries": [ "llm-rl", "post-training", "reward-model" ], "source_provenance": [] }, { "id": "2609.01591", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "StudentSim: Training LLM-based Student Simulators", "track": "post-training", "matched_queries": [ "llm-rl", "process-reward", "reward-model" ], "source_provenance": [] }, { "id": "2609.01587", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "The Structure of Quantization Damage in LLMs: Why the Next Bit Should Be Spent Globally", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.01573", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Scaling Near-Optimal SFT-RL Annotation Budget Allocation from Small to Large LLMs", "track": "post-training", "matched_queries": [ "llm-rl", "post-training", "preference-optimization", "reward-model" ], "source_provenance": [] }, { "id": "2609.01572", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "From Production Traffic to Post-Training: Building a Self-Hosted LLM That Covers the Corporate Request Mix", "track": "post-training", "matched_queries": [ "post-training", "reward-model" ], "source_provenance": [] }, { "id": "2609.01567", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Selective Agent Guidance via Entropy: Learning Autonomous Policies from Imperfect VLM Teachers", "track": "post-training", "matched_queries": [ "llm-rl", "on-policy-distillation", "opd", "reward-model", "test-time-rl" ], "source_provenance": [] }, { "id": "2609.01532", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [ { "source": "Meta Research GitHub", "source_kind": "github-organization", "source_url": "https://api.github.com/orgs/facebookresearch/repos?sort=updated&per_page=100", "organization": "Meta", "relation": "direct", "seed_arxiv_id": null } ] }, { "id": "2609.01422", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "From Rollouts to Recipes: Self-Contained Post-Training for LLMs", "track": "post-training", "matched_queries": [ "on-policy-distillation", "post-training" ], "source_provenance": [] }, { "id": "2609.01409", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "EdiTikZ: Scientific Figure Editing from Revision Trajectories", "track": "post-training", "matched_queries": [ "llm-rl", "reward-model" ], "source_provenance": [] }, { "id": "2609.01687", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Learning Evidence Sufficiency Boundaries for Selective Answering in Grounded Multi-Hop QA", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.01354", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Where the Verifier Fails: A Category-Level Audit of Reward Signals in RLVR", "track": "post-training", "matched_queries": [ "llm-rl", "reward-model" ], "source_provenance": [] }, { "id": "2609.01341", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Probing Factual Knowledge Transfer with Training Data Interventions", "track": "post-training", "matched_queries": [ "reward-model" ], "source_provenance": [] }, { "id": "2609.01325", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "VerTox: Verifiable Reward-Guided Corpus Poisoning Against Neural Ranking Models", "track": "post-training", "matched_queries": [ "llm-rl", "reward-model" ], "source_provenance": [] }, { "id": "2609.01274", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "From Base Rollouts to RL Reasoning: A Budgeted Search Perspective", "track": "post-training", "matched_queries": [ "llm-rl", "reward-model", "test-time-rl" ], "source_provenance": [] }, { "id": "2609.01246", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Ready to Speak: Aligning LLMs for TTS-Friendly Text Generation", "track": "post-training", "matched_queries": [ "preference-optimization", "reward-model" ], "source_provenance": [] }, { "id": "2609.01245", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Explore More, Drift Less: Outcome-Only Reinforcement Learning Can Suffice for Long-Horizon Interactive Agents", "track": "post-training", "matched_queries": [ "test-time-rl" ], "source_provenance": [] }, { "id": "2609.01244", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Post-Training Science for Supervised Fine-Tuning", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.01195", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CaRL-EM: Cost-Aware Reinforcement Learning for Entity Matching with LLMs", "track": "post-training", "matched_queries": [ "llm-rl" ], "source_provenance": [] }, { "id": "2609.01188", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "PersuaRL: Reinforcement Learning-Driven Multi-Expert Selection for Persuasive Dialogue Generation in Insurance", "track": "post-training", "matched_queries": [ "llm-rl" ], "source_provenance": [] }, { "id": "2609.01161", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CopyShield: A Cross-Level Benchmark of Copyright Defenses in LLMs", "track": "post-training", "matched_queries": [ "preference-optimization" ], "source_provenance": [] }, { "id": "2609.01073", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Post-hoc Alignment of LLM-judges to Human Judgment Distribution", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.00925", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Context-Grounding Gains Are Mediated by Pre-existing Machinery: Auditing GRPO, SFT, and DPO", "track": "post-training", "matched_queries": [ "post-training", "reward-model" ], "source_provenance": [] }, { "id": "2609.00833", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Dense Process Supervision for Search Agents via Fact Utility Estimation", "track": "post-training", "matched_queries": [ "llm-rl", "process-reward", "reward-model" ], "source_provenance": [] }, { "id": "2609.00796", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SFAD: Speculative Factuality-Aware Decoding", "track": "post-training", "matched_queries": [ "llm-rl", "post-training", "preference-optimization" ], "source_provenance": [] }, { "id": "2609.00791", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Instella-MoE Technical Report", "track": "post-training", "matched_queries": [ "llm-rl", "on-policy-distillation", "post-training", "preference-optimization" ], "source_provenance": [] }, { "id": "2609.00746", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Text Capability Loss in Vision-Language Adaptation: An Attention-Sink Diagnosis", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.00709", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Controllable Image Captioning with Prompt-Conditioned Scene Rewards", "track": "post-training", "matched_queries": [ "reward-model" ], "source_provenance": [] }, { "id": "2609.00590", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CRAFT: Fine-Tuning Pre-hoc Explainability in AI-native 6G RAN", "track": "post-training", "matched_queries": [ "post-training", "preference-optimization", "process-reward", "reward-model" ], "source_provenance": [] }, { "id": "2609.00576", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Consistency Without Alignment: Item-Sensitive Language Models Indistinguishable From Random", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.00565", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Aligned but Flattened: Analyzing the Trade-off between Cultural Alignment and Diversity in LLMs", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.00550", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Same Semantics, Different Outcome: On the Modality Robustness of Multimodal LLMs under Knowledge Conflict", "track": "post-training", "matched_queries": [ "preference-optimization" ], "source_provenance": [] }, { "id": "2609.00457", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Can LLMs Use Relational Transformer Embeddings?", "track": "post-training", "matched_queries": [ "llm-rl" ], "source_provenance": [] }, { "id": "2609.00444", "key": "gapo", "status": "implemented", "priority": "P1", "title": "Group Adaptive Clipping Policy Optimization", "track": "post-training", "doc": "post-training/2609.00444-gapo/README.md", "matched_queries": [ "llm-rl", "reward-model" ], "source_provenance": [] }, { "id": "2609.00355", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Vision Is Not Overhead: One-Pass Block Drafting for Lossless Speculative Decoding in Vision-Language Models", "track": "post-training", "matched_queries": [ "reward-model" ], "source_provenance": [] }, { "id": "2609.00224", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "QTEA: Ternary LLMs with Sparse Residual Salient Weight and By-Column Optimization", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.00213", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Uncovering and Mitigating Aggregation-Induced Reward Hacking in Multi-Reward Reinforcement Learning", "track": "post-training", "matched_queries": [ "llm-rl", "reward-model" ], "source_provenance": [] }, { "id": "2608.31119", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "PaperGym: Rubric-Centered Evolution for Research-Plan Generation", "track": "post-training", "matched_queries": [ "llm-rl", "reward-model" ], "source_provenance": [] }, { "id": "2608.31079", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Sycophantic Agreement Transfers with Neutral Data via Contrastive Preference Optimization", "track": "post-training", "matched_queries": [ "post-training", "preference-optimization" ], "source_provenance": [] }, { "id": "2608.31046", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Does On-Policy Distillation Really Distill? From Noisy Teacher to Self-Improvement", "track": "post-training", "matched_queries": [ "llm-rl", "on-policy-distillation", "reward-model" ], "source_provenance": [] }, { "id": "2608.31035", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "When Does Predictor-Based RL Align with Human Perception? A Study of Subjective Rewards in Codec-Based Speech Language Models", "track": "post-training", "matched_queries": [ "llm-rl", "post-training", "reward-model" ], "source_provenance": [] }, { "id": "2608.31007", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Augmenting Interviewer Judgments of Patient Experience with Automatic Language Analysis", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2608.30980", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Evaluating and Improving LLM Self-Modeling", "track": "post-training", "matched_queries": [ "llm-rl" ], "source_provenance": [] }, { "id": "2608.30968", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CogEvol: Towards Efficient and Reliable Learning Environment Generation", "track": "post-training", "matched_queries": [ "reward-model" ], "source_provenance": [] }, { "id": "2608.30952", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "One Policy Is Enough: Single-Agent Reinforcement Learning Outperforms Tree Search for Chemistry Tool Learning", "track": "post-training", "matched_queries": [ "llm-rl", "reward-model" ], "source_provenance": [] }, { "id": "2608.30910", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "S3C-LLM: Skill-Code Guided Agentic Language Models for Spectrum-to-Structure Elucidation", "track": "post-training", "matched_queries": [ "llm-rl" ], "source_provenance": [] }, { "id": "2608.30908", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Fine-Tuning Low-Bit Models with Gradient in Quantized Code Space", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.01658", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "PRO-Step: Step-level Process Reward Optimization for Retrieval-Augmented Generation", "track": "post-training", "matched_queries": [ "preference-optimization", "process-reward", "reward-model" ], "source_provenance": [] }, { "id": "2608.30841", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "HSRM: Hidden-State Reward Models for Test-Time Verification", "track": "post-training", "matched_queries": [ "process-reward", "reward-model" ], "source_provenance": [] }, { "id": "2608.30760", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "PRACTICE: From Experience to Expertise in Self-Evolving Embodied Agents", "track": "post-training", "matched_queries": [ "on-policy-distillation", "opd" ], "source_provenance": [] }, { "id": "2608.30731", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Calibrating Small Language Models for Claim Check-Worthiness Detection", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2608.30719", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Mind the Gap: Theory-of-Mind-Grounded Friction for Epistemic Alignment", "track": "post-training", "matched_queries": [ "preference-optimization", "reward-model" ], "source_provenance": [] }, { "id": "2608.30632", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "GMTS: Gradient Magnitude-based Token Selection Improves RLVR Training for LLM Reasoning", "track": "post-training", "matched_queries": [ "llm-rl", "reward-model" ], "source_provenance": [] }, { "id": "2608.30597", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "PLC-DPO: Posterior Label Correction in Noisy and Ambiguous Preference Optimization", "track": "post-training", "matched_queries": [ "preference-optimization" ], "source_provenance": [] }, { "id": "2608.30427", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Ceiling-Clipped Acceptance Histograms Indicate Stranded Speed-up in Block-Diffusion Speculative Decoding", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2608.30406", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Locally-Guided Actor-Critic: Training a Goal-conditioned Actor with a Subgoal-aware Critic", "track": "post-training", "matched_queries": [ "test-time-rl" ], "source_provenance": [] }, { "id": "2608.30399", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SemPOI-RL: Aligning LLM Semantic Reasoning for Interpretable Out-of-Town POI Sequential Generation", "track": "post-training", "matched_queries": [ "llm-rl", "reward-model" ], "source_provenance": [] }, { "id": "2608.30398", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Beyond Polarization: The Generative Constraint of Chain-of-Thought in Pointwise Reranking", "track": "post-training", "matched_queries": [ "llm-rl" ], "source_provenance": [] }, { "id": "2608.30325", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Sequential Trajectories and Simultaneous Blending: Multi-Emotion Modeling for Instruction-Following TTS", "track": "post-training", "matched_queries": [ "post-training", "preference-optimization", "reward-model" ], "source_provenance": [] }, { "id": "2608.30141", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Balancing Privacy, Utility, and Safety in LLM Alignment through Preference Optimization", "track": "post-training", "matched_queries": [ "preference-optimization" ], "source_provenance": [] }, { "id": "2608.30125", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "VIBE: Video Instruction-aligned Background music gEneration", "track": "post-training", "matched_queries": [ "reward-model" ], "source_provenance": [] }, { "id": "2608.30109", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "COGTRL: Training LLMs for Scientific Discovery Assistance using Cognitive Traces via Reinforcement Learning", "track": "post-training", "matched_queries": [ "llm-rl", "preference-optimization" ], "source_provenance": [] } ] }, { "batch": "recent-20260905-agent", "date": "2026-09-05", "scope_kind": "automated-review-closure", "required_tracks": [ "agent" ], "source_artifact": "/tmp/auto-research-discovery-20260905/agent-paper-candidates/paper-candidates-agent.json", "window": { "start": "2026-08-31", "end": "2026-09-05" }, "candidates": [ { "id": "2609.04170", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "A Case Study on Emergent Cheating and Whistleblowing in Autonomous Research Swarms", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04167", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SWE-Gate: Passing Functional Tests Is Not Enough for Software Engineering Agents", "track": "agent", "matched_queries": [ "llm-agent", "software-agent" ], "source_provenance": [] }, { "id": "2609.04159", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SENTINEL-RL: Offloading Topological Reasoning from LLM Agents in the Security Operations Center", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.04135", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "The Natural Language Interaction Protocol and Standard for AI Agents", "track": "agent", "matched_queries": [ "tool-agent" ], "source_provenance": [] }, { "id": "2609.04094", "key": "draco", "status": "implemented", "priority": "P1", "title": "DRACO: Fine-Grained Credit Assignment with Dynamic Rubrics for Long-Horizon Agent Training", "track": "agent", "doc": "agent-research/2609.04094-draco/README.md", "matched_queries": [ "agentic-rl", "software-agent" ], "source_provenance": [] }, { "id": "2609.04075", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "PatchBench: Evaluating AI Agents for Vulnerability Patching", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.04063", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Spurious Advantage Hidden in GRPO", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.04034", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Editable Visual Design", "track": "agent", "matched_queries": [ "agent-planning" ], "source_provenance": [] }, { "id": "2609.03960", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "FiMI Banking: A Sovereign Model for Indian Retail Banking", "track": "agent", "matched_queries": [ "agentic-rl", "tool-agent" ], "source_provenance": [] }, { "id": "2609.03923", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Speak for Me: Giving LLMs the Situational Awareness to Participate in a Meeting", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.03920", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Value-Preserving Architectures for Agentic AI Systems", "track": "agent", "matched_queries": [ "llm-agent", "software-agent" ], "source_provenance": [] }, { "id": "2609.03884", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "A Blind Trust, the Bloody Thrust: When Attacker-Controlled Hook Updates Steer AI Agent Harnesses towards Malicious Behaviors", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.03871", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Bioinfoysis Technical Report", "track": "agent", "matched_queries": [ "agent-memory", "agent-planning", "multi-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2609.03860", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Adapting to Evolving Requirements: Agentic AI for Retail Supply Chain Operations", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.03846", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "EF1-Constrained Nash Social Welfare with Identical Additive Valuations: Complexity, Guarantees, and Experiments", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.03834", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Semantic Bayesian World Models", "track": "agent", "matched_queries": [ "agent-planning", "web-agent" ], "source_provenance": [] }, { "id": "2609.03753", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SimSkill: A Lifelong Learning AI Agent for Autonomous Mastery of Traffic Simulation", "track": "agent", "matched_queries": [ "agent-memory", "llm-agent" ], "source_provenance": [] }, { "id": "2609.03727", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Proactive Service Agents: A Unified Decision Framework, Methods, and Evaluation", "track": "agent", "matched_queries": [ "agent-memory", "agent-planning", "software-agent" ], "source_provenance": [] }, { "id": "2609.03718", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "What Do CAE Simulation Agents Really Need Beyond a Generic Harness?", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2609.03667", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Out-of-Distribution Generalisation with Sequence Models in Offline Multi-Agent Reinforcement Learning", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.03619", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Remember and Reweight: Enhancing Multi-Agent Debate with Experience Memory and Confidence Estimation", "track": "agent", "matched_queries": [ "agent-memory", "multi-agent" ], "source_provenance": [] }, { "id": "2609.03590", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "From Prior-Guided Heuristics to Deployable Agents: Accelerating Demonstration-Driven Reinforcement Learning for Deadline-Constrained Network Control", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.03588", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "KC-Bench: A Dynamic Interactive Benchmark for Evaluating Knowledge Conflicts in LLM Agents", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.03553", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "GPS-Bench: A Governance Policy Benchmark for Automating Policy Analysis", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.03493", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Making Every Tool Call Count: Necessary Tool-Evidence Path Rewards for Agentic Vision-Language Models", "track": "agent", "matched_queries": [ "tool-agent" ], "source_provenance": [] }, { "id": "2609.03483", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Air-Ground Collaborative Vision-and-Language Navigation via Shared Bird's-Eye Maps", "track": "agent", "matched_queries": [ "agent-planning" ], "source_provenance": [] }, { "id": "2609.03467", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "When Users Don't Ask: Benchmarking Context-Driven Memory Retrieval in Conversational Agents", "track": "agent", "matched_queries": [ "agent-memory", "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.03450", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Plan Pointers and Record-Directive Form in Budgeted Verification of Inherited Agent Memory", "track": "agent", "matched_queries": [ "agent-memory", "agent-planning" ], "source_provenance": [] }, { "id": "2609.03423", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "DuplexSpeechBench-IFEval: Evaluating Implicit Instruction Following in Full-Duplex Voice Agents", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.03416", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Dude: A Dual-Detection Multi-Agent System for Paper-Code Discrepancy Detection", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.03414", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "StrixAE: An Intelligent Agent for Audio Enhancement under Complex Distortion Coupling in Real-World Scenarios", "track": "agent", "matched_queries": [ "agent-planning", "agentic-rl" ], "source_provenance": [] }, { "id": "2609.03402", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "A Prompt-Engineering Approach to Develop Scalable, Flexible, and Real-Time Hybrid Micro-Level Personalization in a General Purpose AI Teaching Assistant", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.03366", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Accountable AI with Grounded, Faithful, Consistent, Actionable Rationales: A Case Study in Clinical Trial Matching with VERDICT", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.03340", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Fresh Memory, Stale Plans: Dependency-Scoped Validation for Distributed LLM-Agent Memory", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.03321", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Decoupling Turn-Taking from Semantics: A Decoupled Data Approach for Finite-State-Machine-Based Full-Duplex Dialogue", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.03308", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Risk and Anomaly Identification for Distribution Network Optimal Operation Based on Reinforcement Learning and Uncertainty Quantification", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.03236", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Speculative Macro Commit for Faster Tool-Using Agents", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.03230", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Two Truths and A Lie? Benchmarking Off-the-Shelf LLMs for Requirements Quality Assessment: Performance, False Alarms, and Misses", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.03221", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Counterfactual Fairness Audits of Multi-Step Clinical LLM Agents Require a Measured Per-Action Instability Floor", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.03209", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "MasterControl Seventeen Every Time", "track": "agent", "matched_queries": [ "agent-planning", "tool-agent" ], "source_provenance": [] }, { "id": "2609.03206", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Learning to Zoom Efficiently with a Contrastive Curriculum", "track": "agent", "matched_queries": [ "tool-agent" ], "source_provenance": [] }, { "id": "2609.03201", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "MemoryLACE: Memory Lifecycle-Aware Consolidation and Evidence Retrieval", "track": "agent", "matched_queries": [ "agent-memory", "llm-agent" ], "source_provenance": [] }, { "id": "2609.03028", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Requirements After the First Edit: Mining Late Requirement Emergence and Rework in Real-World Coding-Agent Sessions", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.02805", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Large Language Models (LLMs) for Telecom Root Cause Analysis (RCA): A Structured Reasoning Framework for Evidence-Grounded Diagnosis", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02786", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SafeEvolve: Harness-Policy Co-Evolution from Agent Experience for Safety Alignment", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02783", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "EarlyEval: Cheaper Agent Evaluation via Early Outcome Prediction", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02750", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Bilevel Coordinated Reflection: A Game-Theoretic Approach to Multi-Agent LLM Systems", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02749", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills", "track": "agent", "matched_queries": [ "agent-memory", "agent-planning" ], "source_provenance": [] }, { "id": "2609.02730", "status": "implemented", "priority": "P0", "reason": "纠正摘要级误拒：全文证实生产推荐闭环、量化线上 A/B 与部署配置；同一实现同时覆盖推荐与 Agent 控制器视角。", "key": "coral", "doc": "reproductions/2609.02730-coral/README.md", "full_text_review": { "scope": "full-text", "source_locations": [ "Section 5.1, Table 1", "Section 5.2" ] }, "title": "CORAL: An LLM-Native Harness for Production Recommender Systems", "track": "agent", "matched_queries": [ "agent-memory", "llm-agent" ], "source_provenance": [] }, { "id": "2609.02649", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Loom: Weaving Diagnostic Strands into Free-Text Consensus via Embedding-Space Reweighting", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02580", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Competitive Market Behavior of LLMs", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02566", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Online Reinforcement Learning in the Met Office Unified Model through Distributed Model-Agent Coupling", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.02459", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CivBench: A Long-Horizon Benchmark for Tool-Mediated Agents in Civilization VI", "track": "agent", "matched_queries": [ "agent-planning", "tool-agent" ], "source_provenance": [] }, { "id": "2609.02371", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Diagnosing with Insights: Structured Analysis of Agent Failures via Behavioral Abstractions", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02322", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "What Is Worth Representing? Representational Empowerment for Continual Model Construction", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02272", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "PaperCompiler: Faithful Paper-to-Code Generation via Repository-Level Specification Compilation", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.02264", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Codebook Agent: Amortized Topology Design for LLM Multi-Agent Systems", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02253", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "APEx: Distillation of Agent Procedural Experience for Adaptive Deep Research Question Answering", "track": "agent", "matched_queries": [ "agent-memory", "agentic-rl", "multi-agent" ], "source_provenance": [] }, { "id": "2609.02967", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Privacy-Preserving Topology-Guided Safety for LLM-Based Multi-Agent Systems via Federated Graph Learning", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02250", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "RideSkill: A Hierarchical Algorithm for Generalized Ride Sharing with LLM-Driven Automatic Evolution", "track": "agent", "matched_queries": [ "agentic-rl", "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.02246", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "LLM-as-a-Judge Is Not an Oracle: Why Self-Improving Agents Need Deterministic Guardrails", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02236", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "PGPO: Potential-Guided Policy Optimization for Multi-Turn Agentic Tasks", "track": "agent", "matched_queries": [ "agentic-rl", "llm-agent" ], "source_provenance": [] }, { "id": "2609.02231", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "PhoenixNest-Video: Evidence-Grounded Multimodal Agent Framework for Automated Video Interview Assessment", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.02217", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SkillGLoW: Procedural-Family Skill Consolidation for Self-Improving Agents on Long-Horizon Task Streams", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02964", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "When Optimization Becomes Manipulation: Defending Generative Search against Malicious Generative Engine Optimization", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02170", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "DMRL: Document-Mediated Reinforcement Learning for Skill Optimization in Advertising Recommendation", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.02149", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "OmegaUse-SOP: SOP Engineering for Professional Computer Use from Human Demonstrations", "track": "agent", "matched_queries": [ "llm-agent", "software-agent" ], "source_provenance": [] }, { "id": "2609.02122", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "AI agents reshape consensus formation in human groups", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02106", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Git4Data: Database-Native Version Control for AI Agents", "track": "agent", "matched_queries": [ "llm-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2609.02094", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "MASkills: Continual Skills Optimization for Multi-Agent LLM Systems", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02092", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Beyond Outcome Gaps: Process-Aware Fairness Diagnosis for LLM-based Multi-Agent Decision Systems", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02067", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "ToolGate: An Executable Acceptance Pipeline for Tool-Dependent Scientific Benchmark Construction", "track": "agent", "matched_queries": [ "software-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2609.02056", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "HyGRAIL: Cost-Aware and Evidence-Grounded Scientific Hypothesis Discovery over Knowledge Graphs", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.02054", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "A Tri-Agent Framework for Evaluating and Aligning Question Clarification Capabilities of Large Language Models", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.02042", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Act More, Decide Less: Skill-Guided Adaptive Action Chunking for Long-Horizon LLM Agents", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.01985", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "When Agents Implement Systems: A Case Study in Defects, Detection, and Evaluation Rigor", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01971", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "NS-Copilot: An LLM-Driven Agent System for Autonomous Neuroscience Analysis", "track": "agent", "matched_queries": [ "agent-memory", "agent-planning", "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.01918", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Grounded, Compute-Efficient LLM Policy Agents for Energy-Poverty Equity in Physically-Constrained Peer-to-Peer Energy Markets", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01873", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Epistemic Sybil Resistance: Multiplying AI Agents Without Multiplying Evidence", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01865", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "ExecRetrieval: Measuring the Functional-Correctness Gap in Code-Embedding Retrieval", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.01861", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Belief-Calibrated Optimization: An Explicit World Model for Agentic Optimization", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01852", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "The Memory Trust Gap: Capability-Dependent Failures in Persistent-Memory Agents", "track": "agent", "matched_queries": [ "agent-memory" ], "source_provenance": [] }, { "id": "2609.01836", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Agent Memory Is a Surface for Endogenous Authorization Laundering", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01779", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Agents That Model Agents: Five Principles Toward a Theory of Mind for 6G Networks", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.01736", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Harness Engineering in LLM Tool Use via Agent-Native Reusable Tool Primitives", "track": "agent", "matched_queries": [ "agent-planning", "llm-agent", "multi-agent", "software-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2609.01603", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Efficient SWE Agent Benchmarking via Trajectory-Aware Evaluation", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.01597", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "The Rise of Verbal Reinforcement Learning", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.01567", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Selective Agent Guidance via Entropy: Learning Autonomous Policies from Imperfect VLM Teachers", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.01702", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "RosettaBitcoin: An Artifact-Backed Experience Report on Verification Infrastructure for Agent-Assisted Consensus Validators", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.01556", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Retrieved but not ranked: surface-form bias in structural retrieval, from mathematics to agent trajectories", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01549", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "NashDreamer: Model-Based Reinforcement Learning for Zero-Sum Imperfect-Information Games", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.01532", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall", "track": "agent", "matched_queries": [], "source_provenance": [ { "source": "Meta Research GitHub", "source_kind": "github-organization", "source_url": "https://api.github.com/orgs/facebookresearch/repos?sort=updated&per_page=100", "organization": "Meta", "relation": "direct", "seed_arxiv_id": null } ] }, { "id": "2609.01526", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "EvoSCM: Scientific Belief Revision Through Causal Model Evolution and Experimentation", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01519", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "When Guardrails Look Effective: Construct Validity Failures in LLM Agent Commerce Evaluation", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01693", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Public-Sharing Labels and Verbatim Field Egress in an MCP-to-A2A Agent Configuration: A Controlled Multi-Model Study", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01491", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "GlossoGen: Emergent Language in Complex Multi-Agent LLM Interactions", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01481", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Harness-of-Harness: Multi-Day Autonomous Software Development with Continual Improvement", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01466", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Parsing the Stream: A Live Trace Model for Long-Horizon Agents and Their Observers", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01437", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness?", "track": "agent", "matched_queries": [ "llm-agent", "software-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2609.01418", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Provably Safe Sim-to-Real Transfer", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.01409", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "EdiTikZ: Scientific Figure Editing from Revision Trajectories", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.01404", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Evaluating Multimodal LLMs as Generalist Vision-Language-Action Agents for Drone Control: Commanding, Approaching, Tracking and Searching", "track": "agent", "matched_queries": [ "agent-planning", "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.01383", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "InSight: A Benchmark for Agentic Claim Verification in Interactive Visualizations", "track": "agent", "matched_queries": [ "web-agent" ], "source_provenance": [] }, { "id": "2609.01360", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "EDGE: Error Dependency Graph-Guided Multi-Error Attribution in Multi-Agent LLM Systems", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.01337", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "LEAP: Likelihood Elicitation and Aggregation for LLM-based Probabilistic Forecasting", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2609.01335", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Bandits in Prod: Hyperparameter Optimization at Inference Time", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01294", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Explore Before Committing: Hypothesis-Guided Search for Deep Research Agents", "track": "agent", "matched_queries": [ "tool-agent" ], "source_provenance": [] }, { "id": "2609.01281", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "EmbodiedSkills: A Unified Framework for Orchestrating, Training, and Deploying VLA Agents", "track": "agent", "matched_queries": [ "agent-memory", "agent-planning" ], "source_provenance": [] }, { "id": "2609.01272", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Making Prospective Memory SLM-Shaped: Typed Intention Stores for Small-Model Agents", "track": "agent", "matched_queries": [ "agent-memory", "llm-agent" ], "source_provenance": [] }, { "id": "2609.01271", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "What Does an Agentic Software Engineering Benchmark Measure? Profiling Task Demands and Agent Behaviour Beyond What Category Labels Reveal", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.01245", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Explore More, Drift Less: Outcome-Only Reinforcement Learning Can Suffice for Long-Horizon Interactive Agents", "track": "agent", "matched_queries": [ "agentic-rl", "llm-agent" ], "source_provenance": [] }, { "id": "2609.01236", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Continuous Autonomous Refactoring: A Research Roadmap for AI-Driven Code Quality Maintenance", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.01188", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "PersuaRL: Reinforcement Learning-Driven Multi-Expert Selection for Persuasive Dialogue Generation in Insurance", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.01111", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "ClinTraceBench: Source-Verifiable Longitudinal Clinical Reasoning over EHR-Derived Dialogues", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01095", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Reliable LLM-Generated Programs for High-Energy Physics Experiments through Graph-Grounded Software Knowledge", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.01058", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "ARISE-RL: Agentic Rubric-Grounded Iterative Self-Evolution with Reinforcement Learning", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.01056", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "WorldBench: Culturally Grounded Benchmark for Multilingual Agents", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.01045", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "AgentFactory: Towards Automated Agentic System Design and Optimization", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01038", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Data-Driven Persona-Conditioned Agents for A/B Test Simulation", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01035", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Spawn Freely, Act Sparingly: Progressive Risk Vesting for Recursive LLM-Agent Trees", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.00949", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Calibration is the Bottleneck: An Action-Class Diagnostic of Multi-Turn Tool-Calling", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.00940", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "A Dataset for Modeling Iterative Problem-Solving", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.00859", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Reinforcement Learning Enhanced LLM Agents for Complex Vehicle Routing Problems", "track": "agent", "matched_queries": [ "agentic-rl", "llm-agent" ], "source_provenance": [] }, { "id": "2609.00833", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Dense Process Supervision for Search Agents via Fact Utility Estimation", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.00813", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "One Policy, Any Budget: Internalizing Budget-Aware Search via Reinforcement Learning", "track": "agent", "matched_queries": [ "agentic-rl", "llm-agent" ], "source_provenance": [] }, { "id": "2609.00802", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "TEIDAN: A Multilingual Multiparty Dialogue Corpus", "track": "agent", "matched_queries": [ "agent-planning" ], "source_provenance": [] }, { "id": "2609.00795", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Agentic programs: an emerging form of scientific software in computational materials science", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.00759", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Compile, Don't Memorize: A Context Compilation Architecture (CCA) for In-Context Learning", "track": "agent", "matched_queries": [ "agent-planning", "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.00749", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "ContextPipe: Database-Inspired Context Assembly for Long-Horizon Agents", "track": "agent", "matched_queries": [ "agent-planning", "llm-agent" ], "source_provenance": [] }, { "id": "2609.00731", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Agentic Empirical Asset Pricing: Methodological Foundations", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.00714", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "ChatDev 2.0: A No-Code Multi-Agent Platform for Developing Everything", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.00652", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Self-Reports Are Not Verification: Environment-Grounded Auditing of LLM Operators in Evolutionary Search", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.00621", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Control-Data Flow Separation: Stable Prompt Optimization in Multi-Agent LLMs", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.00608", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Investigating Assistant Bias in LLM User Simulators Using a Role Vector", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.00595", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SoK: When Safe Agents Fail Together: The Security of Multi Agent LLM Systems", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.00577", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "GeoPAR: Large-Scale Multi-Agent Combinatorial Optimization with Geometry-Guided Parallel Autoregressive Learning", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.00570", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "VoiceLongMemEval: Do Assistants Remember How You Sounded?", "track": "agent", "matched_queries": [ "multi-agent" ], "source_provenance": [] }, { "id": "2609.00568", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "WiseSpec: Requirements-Driven Agents for Code Generation", "track": "agent", "matched_queries": [ "llm-agent", "software-agent" ], "source_provenance": [] }, { "id": "2609.00549", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Skill Following: Evaluating Actual Skill Use in Retrieval-Enabled LLM Agents", "track": "agent", "matched_queries": [ "llm-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2609.00546", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Runtime-Independent Persistent Agents: Preserving Identity, Memory, and Code Across Models, Harnesses, and Servers", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.00524", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Are We There Yet? Assessing Computer-Use Agents for Blind Users' Accessible Interaction with Desktop Applications", "track": "agent", "matched_queries": [ "agent-planning" ], "source_provenance": [] }, { "id": "2609.00504", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Independent Reinforcement Learning in Discounted Markov Games", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.00474", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Exploring Collaboration between a language and a non-language agent", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.00455", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Towards a Belief-Based World Model for LLM Agents", "track": "agent", "matched_queries": [ "agent-planning", "llm-agent" ], "source_provenance": [] }, { "id": "2609.00434", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SAGE: State-Grounded, Abstention-Aware Evaluation of Task-Oriented Dialogue Agents", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.00384", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "RestoreBench: Can AI Agents Restore Power Flow Convergence?", "track": "agent", "matched_queries": [ "agent-planning", "llm-agent", "multi-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2609.00377", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "FoldingAgent: Inferring Parametric Origami Procedures from Demonstration Videos", "track": "agent", "matched_queries": [ "agent-planning" ], "source_provenance": [] }, { "id": "2609.00296", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Toward Workflow-Aware Benchmarking for Healthcare NLP Agents", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.00275", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "The Irreversibility Budget: Fleet-Level Risk Accounting and Admission Control for Agent Operating Systems", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.01660", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "How Fast Do Agents Rot? An Empirical Study of Long-Horizon Degradation in LLM Agents for Production Decision-Making", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2609.00267", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Delegation Without Trust: An Empirical Gap Analysis of Identity, Authorization, and Runtime Governance in Multi-Agent LLM Systems", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2609.00256", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "NSIDDx: A Design Framework for Neuro-Symbolic, Practitioner-First Differential Diagnosis in Low-Resource Settings", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.00252", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Spec-Driven Development for Agentic Software Engineering: Harnessing Human-Agent Teamwork", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.00243", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Invalidation Contracts for Cross-Episode Agent Memory", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.00237", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Learning What to Retain: Gated-Memory Routing for Efficient Collaboration in Multi-Agent LLM Systems", "track": "agent", "matched_queries": [ "agent-memory", "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.00227", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Don't Let the Model Write the YAML: Deterministic, Minimal-Diff GitOps Remediation from LLM-Proposed Field Changes", "track": "agent", "matched_queries": [ "llm-agent", "software-agent" ], "source_provenance": [] }, { "id": "2609.00206", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Distributed Implicit Harm: A Compositional Safety Blind Spot in MLLM-Based Video Moderation", "track": "agent", "matched_queries": [ "multi-agent" ], "source_provenance": [] }, { "id": "2609.00193", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Provably Efficient Federated Reinforcement Learning with Linear Function Approximation and Logarithmic Communication Cost", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2608.31128", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "DIASENTINEL: An Auditable Multi-Agent System for Guideline-Grounded Diabetes Risk Screening", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2608.31111", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Aspire: Can Models Self-Evolve from Vague Goals?", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2608.31100", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement?", "track": "agent", "matched_queries": [ "agent-memory", "llm-agent" ], "source_provenance": [] }, { "id": "2609.00103", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Good Memory Has ECC: Evaluating the Memory of Vision-Language Models Beyond Accuracy", "track": "agent", "matched_queries": [ "agent-memory", "llm-agent" ], "source_provenance": [] }, { "id": "2608.31082", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Token-Efficient Data Reasoning Agents via Adaptive Structuring of Unstructured Data", "track": "agent", "matched_queries": [ "llm-agent", "web-agent" ], "source_provenance": [] }, { "id": "2608.31077", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Reconciling Process Supervision with Outcome-Based Credit in Agentic Policy Optimization", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2608.31076", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Learning to Evaluate Before Improving: Automatic Rubric Induction for Automatic Research Agents", "track": "agent", "matched_queries": [ "llm-agent", "software-agent" ], "source_provenance": [] }, { "id": "2608.31075", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Scaling Large Reasoning Models beyond Human Supervision: A Path toward Superintelligence", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2608.30963", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "A Universal Context-Reuse Layer for Cross-Model KV Sharing", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2608.30952", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "One Policy Is Enough: Single-Agent Reinforcement Learning Outperforms Tree Search for Chemistry Tool Learning", "track": "agent", "matched_queries": [ "agentic-rl", "tool-agent" ], "source_provenance": [] }, { "id": "2608.30948", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Detecting AI Impostors: How Do Middle Schoolers Identify LLM Agents in a Live Collaborative Setting?", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2608.30935", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "LightNav-0: Eliciting VLM Spatial Intelligence for Generalist Embodied Navigation", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2608.30924", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "TRIPPULSE: Multi-Agent Travel Planning with Review-Grounded Reasoning", "track": "agent", "matched_queries": [ "agent-planning", "llm-agent" ], "source_provenance": [] }, { "id": "2608.30910", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "S3C-LLM: Skill-Code Guided Agentic Language Models for Spectrum-to-Structure Elucidation", "track": "agent", "matched_queries": [ "agentic-rl", "llm-agent" ], "source_provenance": [] }, { "id": "2608.30803", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Schwarz: Solver-Aware Agentic Program Verification", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2608.30754", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CLIN: an Objective Framework for Evaluating Creativity in Short Persian Literary Text", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2608.30730", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "E-Commerce Bench: Evaluating LLM Agents on Long-Horizon Autonomous Business Operation", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2608.30724", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "BAITBENCH: Measuring Agent Reward Hacking with Optional Shortcuts Planted in ML Tasks", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2608.30716", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SocialReasonBench: A Video-QA Benchmark for Social Reasoning with Counterfactual Narrative Videos", "track": "agent", "matched_queries": [ "multi-agent" ], "source_provenance": [] }, { "id": "2608.30701", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "A Phased Workflow for Operating LLM-Based Coding Agents", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2608.30685", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "ATLAS: Dual-Horizon Diagnostic Evaluation for Industrial Tool-Use Agents", "track": "agent", "matched_queries": [ "llm-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2608.30676", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "MedAgent-R1: Faithfulness-Aware Reinforcement Learning for Evidence-Grounded Medical Reasoning", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2608.30672", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "HiRS-Agent: A Hierarchical Multi-Agent System for Reliable Long-Horizon Remote Sensing Task Solving", "track": "agent", "matched_queries": [ "agentic-rl", "multi-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2608.30661", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SwarmBench: Can Large Language Models Act as Agent Swarm Orchestrators?", "track": "agent", "matched_queries": [ "multi-agent" ], "source_provenance": [] }, { "id": "2608.30659", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "LLM-based Hardware Development with Hierarchical IRs and End-to-End Multi-Agent Workflow", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2608.30650", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Geometry of Divergence: Tracking Hidden-State Trajectories for Adaptive Multi-Turn Reasoning", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2608.30572", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Practical Implementation Report on Introducing Spec-Driven Development Using AI Agents in Software Development PBL", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.00081", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Framework and Benchmark for Code-Driven Agentic Testing in Web Development", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2608.30498", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CM2: Multimodal Cultural Reasoning via an Integrated Multi-Agent Framework", "track": "agent", "matched_queries": [ "multi-agent" ], "source_provenance": [] }, { "id": "2608.30478", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Agents in the Large: Perception-Centered Architecture for Persistent Agents", "track": "agent", "matched_queries": [ "agent-memory", "tool-agent" ], "source_provenance": [] }, { "id": "2608.30429", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "EvoSkill Injection: Red-Teaming Autonomous Skill Generation and Evolution in Self-Evolving Agents", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2608.30428", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Lies We Can See: Joint Verbal and Non-Verbal Deception by VLM Agents in Embodied Social Interactions", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.00077", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Beneath the Diff: Diagnosing and Mitigating Algorithmic Mode Collapse in Code-Level Autonomous Research Loops", "track": "agent", "matched_queries": [ "llm-agent", "software-agent" ], "source_provenance": [] }, { "id": "2608.30406", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Locally-Guided Actor-Critic: Training a Goal-conditioned Actor with a Subgoal-aware Critic", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2608.30396", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Scaffolding Foundation Models into Physical-World Agents Pushes the Frontier of Long-Horizon Navigation", "track": "agent", "matched_queries": [ "agent-memory", "agent-planning" ], "source_provenance": [] }, { "id": "2608.30391", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Using Grounded Theory for Agent Behavior Analysis at Scale", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2608.30373", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Beyond Consensus: Downward Bias and Role Asymmetry in Multi-Agent LLM Judges for Subjective Evaluation", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2608.30362", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Will the User Ever Know? Covert Indirect Prompt Injection Attacks on Tool-Using LLM Agents", "track": "agent", "matched_queries": [ "llm-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2609.00072", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Can MCP Clients Decide What to Do After Failure? A Result-Only Actionability Audit", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2608.30322", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Ignorance or Incompetence? Constructing Knowledge-Gated, Verifiable Tasks for LLM Agents", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2608.30303", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Lazy Grounding: Attacking Search Agents with Factual Evidence", "track": "agent", "matched_queries": [ "web-agent" ], "source_provenance": [] }, { "id": "2608.30300", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Update from Hell: Can Coding Agents Survive Hidden Breakage in Dependency Upgrades?", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2608.30277", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SimCRAFT: Distilling Remote Sensing Agents via Synthetic Trajectories and Contextual Retrieval-Augmented Fine-Tuning", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2608.30207", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "SIR: Self-improving Red-teaming for Compute Use Agents", "track": "agent", "matched_queries": [ "llm-agent", "web-agent" ], "source_provenance": [] }, { "id": "2608.30192", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "FaVOR: LLM-Based Agentic Framework for Factor Mining via Empirical Validation", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2608.30188", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "GPAgentBench-2K: Benchmarking Large Language Model Agents in Complex Clinical Action Space", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2608.30181", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "A.X K2 Technical Report", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2608.30162", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Reinforcement Learning for Symbolic Equation Solving", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2608.30147", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "CAST: Critique-Aware Supervision for Training Reliable Long-Horizon Tool-Calling Agents", "track": "agent", "matched_queries": [ "llm-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2608.30110", "status": "rejected", "priority": "P2", "reason": "高召回标题命中；经摘要与可执行性复核，不满足本批工业线上证据或独立公开复现门槛。", "title": "Can LLMs Take the Pulse of the Economy? A Real-Time Evaluation of LLM Nowcasts on Macroeconomic Indicators", "track": "agent", "matched_queries": [ "llm-agent", "web-agent" ], "source_provenance": [] } ] }, { "batch": "recent-20260907-recommendation", "date": "2026-09-07", "scope_kind": "automated-review-closure", "required_tracks": [ "recommendation" ], "source_artifact": "/tmp/scan-20260907-recommendation.json", "window": { "start": "2026-09-03", "end": "2026-09-07" }, "candidates": [ { "id": "2609.05363", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Distill Globally, Adapt Locally: Reasoning Distillation and Product-Type Test-Time Training for Scalable Trade-Up Recommendation", "track": "recommendation", "matched_queries": [ "generative-recommendation" ], "source_provenance": [] }, { "id": "2609.05110", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Embedding Surgery: Localized Updates for Adaptive Ranking Correction in Dense Retrieval", "track": "recommendation", "matched_queries": [ "generative-recommendation", "recsys-general" ], "source_provenance": [] }, { "id": "2609.05063", "key": "allecompanion", "status": "implemented", "priority": "P0", "doc": "reproductions/2609.05063-allecompanion/README.md", "title": "Beyond Co-purchase Relation: Evolution of Complementary Recommendations at Allegro", "track": "recommendation", "matched_queries": [ "deployment-evidence", "generative-recommendation", "llm-recommendation", "recsys-general" ], "source_provenance": [] }, { "id": "2609.05059", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Repeated Queries Exhaust an LLM's Brand Recommendations but Not Its Sources", "track": "recommendation", "matched_queries": [ "llm-recommendation" ], "source_provenance": [] }, { "id": "2609.04882", "key": "atomrec", "status": "implemented", "priority": "P1", "doc": "agent-research/2609.04882-atomrec/README.md", "title": "AtomRec: Evolving Atomic Memory for Agentic Recommendation", "track": "recommendation", "matched_queries": [ "recsys-general" ], "source_provenance": [] }, { "id": "2609.04862", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Personalized Task Dependency Graphs for Mitigating Signal Erosion in Multi-Task Recommendation", "track": "recommendation", "matched_queries": [ "industrial-ranking", "production-evidence" ], "source_provenance": [] }, { "id": "2609.04574", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "MURAL: Multimodal Uncertainty-aware Recommendation via Adaptive edge Learning", "track": "recommendation", "matched_queries": [ "generative-recommendation", "priority-org-amazon", "priority-org-tiktok" ], "source_provenance": [] }, { "id": "2609.02928", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Towards Scaling Reinforcement Learning to Massive Populations: Learning Mean-Field Representations", "track": "recommendation", "matched_queries": [ "recsys-general" ], "source_provenance": [] }, { "id": "2609.00014", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Behaviorally Grounded User Profiles from the Wild for Personalized Alignment and Multi-Perspective Reasoning", "track": "recommendation", "matched_queries": [ "llm-recommendation" ], "source_provenance": [] }, { "id": "2609.01628", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Marginal Expected Revenue for Jointly Ranking Auction and Fixed-Price Listings in E-Commerce Sponsored Search", "track": "recommendation", "matched_queries": [ "search-ranking" ], "source_provenance": [] }, { "id": "2609.01627", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "The Utility of LLMs in Recommender Systems Explanation Evaluation", "track": "recommendation", "matched_queries": [ "generative-recommendation", "llm-recommendation", "recsys-general" ], "source_provenance": [] } ] }, { "batch": "recent-20260907-foundation-model", "date": "2026-09-07", "scope_kind": "automated-review-closure", "required_tracks": [ "foundation-model" ], "source_artifact": "/tmp/scan-20260907-foundation-model.json", "window": { "start": "2026-09-03", "end": "2026-09-07" }, "candidates": [ { "id": "2609.05403", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "RegionFed: Federated Learning for Personalized Query Understanding in Heterogeneous Retail Environments", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.05401", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Same Trajectory, Contradictory Rewards (ROBORMBENCH): Paraphrase Fragility in Vision Language Reward Models", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.05388", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Think-Verify-Revise: Neuro-Symbolic Visual Reasoning with Vision-Language Models and Dynamic Logic Tensor Networks", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.05369", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Towards Neuro-Symbolic Procedural Reasoning for Long-Horizon Vision-Language-Action Manipulation", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.05363", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Distill Globally, Adapt Locally: Reasoning Distillation and Product-Type Test-Time Training for Scalable Trade-Up Recommendation", "track": "foundation-model", "matched_queries": [ "efficient-inference" ], "source_provenance": [] }, { "id": "2609.05339", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Does Your Agent's Memory Survive a Model Upgrade? A Controlled Study of Memory Portability", "track": "foundation-model", "matched_queries": [ "long-context", "model-compression" ], "source_provenance": [] }, { "id": "2609.05324", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "RoboSPA: Can VLA Models Go Beyond Simple Scenes and Short-Horizon Tasks?", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.05314", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Large Language Models for HVAC Operations in Building Energy Systems: A Critical Review of Methods, Applications, and Deployment Readiness", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.05239", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Cross-Domain Tracker Adaptation Without Target-Domain Labels via Vision-Language Agents", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.05235", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "PRICE: A Systematic Study of LLM Adaptation Choices for Bitcoin Price Forecasting", "track": "foundation-model", "matched_queries": [ "efficient-inference", "pretraining-data" ], "source_provenance": [] }, { "id": "2609.05224", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "First Things First: Teaching LLM-Based Agents to Prioritize Must-Haves before Nice-to-Haves", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.05152", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Compression Beyond the Uncompressed: A Two-Stage Training Recipe for Soft Context Compression in RAG", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2609.05151", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Large Language Models with At Most One Spike per Neuron", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.05149", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "From Vision to Language: Investigating Causal Information Flow in Multimodal Decision-Making", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.05063", "key": "allecompanion", "status": "implemented", "priority": "P0", "doc": "reproductions/2609.05063-allecompanion/README.md", "title": "Beyond Co-purchase Relation: Evolution of Complementary Recommendations at Allegro", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.05022", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "MoirfEolas and CríochScore: Developing Resources for and the Evaluation of Tokenization Alignment with Irish Morphology", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2609.04971", "key": "beaconkv", "status": "implemented", "priority": "P1", "doc": "reproductions/2609.04971-beaconkv/README.md", "title": "BeaconKV: Key-Value Cache Compression Guided by Beacon Queries for Efficient Large Reasoning Model Inference", "track": "foundation-model", "matched_queries": [ "long-context", "model-compression" ], "source_provenance": [] }, { "id": "2609.04962", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Why We Care About Understanding: Competence through Predictive Compression", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2609.04959", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Discourse Dependency: A Continuous Criterion for Translation Difficulty", "track": "foundation-model", "matched_queries": [ "long-context" ], "source_provenance": [] }, { "id": "2609.04942", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Learning 3D Editing without Paired Supervision via Generative Prior Distillation", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.04852", "key": "kvmem", "status": "implemented", "priority": "P1", "doc": "reproductions/2609.04852-kvmem/README.md", "title": "KVMem: Virtualizing Million-Token Agent Workspaces on a Consumer GPU", "track": "foundation-model", "matched_queries": [ "efficient-inference" ], "source_provenance": [] }, { "id": "2609.04851", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Coupled Control and Wireless World Models for Resilient Remote Robotic Control", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.04842", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "MMTClinic: Multimodal, Multilingual Time Series Question Answering and Reasoning Benchmark for Clinical Domain", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2609.04781", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "CoMLP: Cooperatively-Gated MLPs for Fine-Grained Cross-Modal Information Fusion in Medical Image Segmentation", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.04741", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Where to Look Matters: Learning Influential Views for VLM-based 3D Visual Grounding", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.04721", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Locating and Steering Refusal Beyond Attention", "track": "foundation-model", "matched_queries": [ "llm-architecture", "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.04720", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Knowing What Not to Answer: Selective Non-Compliance in Vision-Language Models", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.04689", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Retinal OCTA Phenotyping with LLM Reporting for Alzheimer's Disease", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.04672", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "WEECFP-SuRGE: Wide Embedded Extended Connectivity Fingerprint with Substructure Rotary Graph-distance Encoding", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.04646", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Importance-Aware Low-Rank Distillation of Diffusion Transformers", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.04645", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Latent-Aligned Reasoning for Multimodal Recommendation", "track": "foundation-model", "matched_queries": [ "llm-architecture", "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.04637", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Tracing Audio Grounding and Answer Selection in Audio LLMs", "track": "foundation-model", "matched_queries": [ "pretraining-data" ], "source_provenance": [] }, { "id": "2609.04598", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "PetQA: Benchmarking Veterinary Knowledge and Clinical Reasoning", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.04550", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "VISTA: Dense Multi-Label Classroom Coding with Vision-Language Models", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.04511", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "LentEx: Generalizable Latent Entity Extraction via Synthetic Data and Instruction-Tuned LLMs", "track": "foundation-model", "matched_queries": [ "efficient-inference" ], "source_provenance": [] }, { "id": "2609.04489", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Towards Understanding Pause Token Fine-Tuning Dynamics: A Mode Retention Perspective", "track": "foundation-model", "matched_queries": [ "model-compression" ], "source_provenance": [] }, { "id": "2609.04486", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Uncertainty Signals for Network Intent Translation: Risk Ranking and Ambiguity Localization", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.04392", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "FAVE: Foveated Adaptive Visual Encoding for Efficient Fine-Grained Visual Understanding", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.04369", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "AdaptVPR: Route-Aware Hard Positive Generation for Robust Visual Place Recognition", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.04355", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "VLA-Precision: Asymmetric Co-Bootstrapping for Efficient Real-World Online RL of Vision-Language-Action Models", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.04336", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "MedProb: Probing Internal Representations of Vision-Language Models for Medical Question Answering", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.04305", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "TNFlow: Amortized Posterior Inference for Trans-Neptunian Object Surface Composition", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.04303", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Abstraction Agent", "track": "foundation-model", "matched_queries": [ "pretraining-data" ], "source_provenance": [] }, { "id": "2609.04289", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Memory as transformation: LETHE, a self-referential gan-inspired architecture", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.04281", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "When Seeing Overrides Knowing: Visual Dominance and Deferral-Based Method for Personalized Safety in VLMs", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.04277", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "FailureSpot: Label-Efficient Timestamp-Level Failure Detection for Vision-Language-Action Models", "track": "foundation-model", "matched_queries": [ "vision-language" ], "source_provenance": [] }, { "id": "2609.04276", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "FailSAE: Towards Interpretable Failure Prediction for Vision-Language Models via Sparse Autoencoders", "track": "foundation-model", "matched_queries": [ "multimodal-llm", "vision-language" ], "source_provenance": [] }, { "id": "2609.04266", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Compute-in-Memory Attention: A Time-Domain Analog Softmax Circuit with RC-Tunable Temperature", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.04263", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Quality Recovery for Quantized KV Caches via Low-Rank Attention Adaptation", "track": "foundation-model", "matched_queries": [ "llm-architecture", "long-context" ], "source_provenance": [] }, { "id": "2609.04261", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Self-Supervised Pretraining of Molecular Graph Encoders with LeJEPA", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] }, { "id": "2609.00068", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Life Operators: a self-evolving framework for multiscale life modelling", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2609.00067", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Do Multimodal LLMs See Before They Read? Diagnosing Contextual Sycophancy", "track": "foundation-model", "matched_queries": [ "multimodal-llm" ], "source_provenance": [] }, { "id": "2609.00051", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "From Detection to Refusal: Safer LLMs via Circuit-Guided Weight Scaling", "track": "foundation-model", "matched_queries": [ "llm-architecture" ], "source_provenance": [] }, { "id": "2609.00050", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Towards Agentic Cloud Engineering: Graph and Loop Engineering with a Zero-Trust Agent Harness", "track": "foundation-model", "matched_queries": [ "transformer-architecture" ], "source_provenance": [] } ] }, { "batch": "recent-20260907-post-training", "date": "2026-09-07", "scope_kind": "automated-review-closure", "required_tracks": [ "post-training" ], "source_artifact": "/tmp/scan-20260907-post-training.json", "window": { "start": "2026-09-03", "end": "2026-09-07" }, "candidates": [ { "id": "2609.05401", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Same Trajectory, Contradictory Rewards (ROBORMBENCH): Paraphrase Fragility in Vision Language Reward Models", "track": "post-training", "matched_queries": [ "reward-model" ], "source_provenance": [] }, { "id": "2609.05314", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Large Language Models for HVAC Operations in Building Energy Systems: A Critical Review of Methods, Applications, and Deployment Readiness", "track": "post-training", "matched_queries": [ "llm-rl" ], "source_provenance": [] }, { "id": "2609.05221", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "A Verifier-Guided Explainable Reasoning Framework with Gold-Anchored QLoRA, Task-Aware Mixture-of-Experts, and Group-Relative RLVR", "track": "post-training", "matched_queries": [ "reward-model" ], "source_provenance": [] }, { "id": "2609.05189", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Can Large Language Models Anticipate Behavioral Responses to Social Policies? A Case of Pension Enrollment Prediction among China's Flexible Workers", "track": "post-training", "matched_queries": [ "on-policy-distillation" ], "source_provenance": [] }, { "id": "2609.05152", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Compression Beyond the Uncompressed: A Two-Stage Training Recipe for Soft Context Compression in RAG", "track": "post-training", "matched_queries": [ "llm-rl" ], "source_provenance": [] }, { "id": "2609.05079", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "TruthInsightBench: An Evidence-Grounded Benchmark for Automated Evaluation of Open-Ended Scientific Discovery Agents", "track": "post-training", "matched_queries": [ "reward-model" ], "source_provenance": [] }, { "id": "2609.04895", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Cache-Aware Joint Router Adaptation for Memory-Efficient MoE Inference", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.04841", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "MABPD: Multi-Agent Bias Probing & Detection via Structured Argument Debate", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.04824", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Generating Constructive Feedback on Stories via Reinforcement Learning", "track": "post-training", "matched_queries": [ "llm-rl", "reward-model" ], "source_provenance": [] }, { "id": "2609.04823", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Reinforcement Learning for improving Large Language Models' Catalan text simplification capabilities", "track": "post-training", "matched_queries": [ "llm-rl", "post-training", "process-reward", "reward-model" ], "source_provenance": [] }, { "id": "2609.04735", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Training Large Language Models for Small-Molecule Design with Synthetic Task Scaling", "track": "post-training", "matched_queries": [ "llm-rl", "post-training", "reward-model" ], "source_provenance": [] }, { "id": "2609.04648", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "ConsensusBench: Benchmark of Consensus Nodes for LLM Reasoning via Outcome Reward Densifying", "track": "post-training", "matched_queries": [ "llm-rl", "process-reward", "reward-model" ], "source_provenance": [] }, { "id": "2609.04565", "key": "sparse-opd", "status": "implemented", "priority": "P0", "doc": "post-training/2609.04565-sparse-opd/README.md", "title": "Extremely Sparse Supervision Incentivizes Reasoning Ability", "track": "post-training", "matched_queries": [ "llm-rl", "on-policy-distillation", "post-training", "process-reward", "reward-model" ], "source_provenance": [] }, { "id": "2609.04485", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Cultural Misalignment in Large Language Models: Detection, Measurement, and Mitigation Through Targeted Fine-Tuning", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.04409", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "A Systematic Evaluation of Cross-Lingual Consistency Enhancement Methods in Multilingual Language Models", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.04355", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "VLA-Precision: Asymmetric Co-Bootstrapping for Efficient Real-World Online RL of Vision-Language-Action Models", "track": "post-training", "matched_queries": [ "llm-rl", "post-training" ], "source_provenance": [] }, { "id": "2609.04344", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "SharedSAE: One Feature Dictionary Across Language Models", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.00058", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "CUDA-Harness: Harnessing Agentic CUDA Kernel Generation and Optimization from Natural Language", "track": "post-training", "matched_queries": [ "reward-model" ], "source_provenance": [] }, { "id": "2609.00057", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "ValueGraph: Value-Signal Guided Graph Pre-training for Contextualized User Representation", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.00052", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "AgentProv: Auditing Agentic LLM API Providers via Tool-use Policy Probes", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.00049", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "REAL-Q: E2E LLM Quantization via Dynamic Gradient Descent", "track": "post-training", "matched_queries": [ "post-training" ], "source_provenance": [] }, { "id": "2609.00028", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "UI-Venus-2 Technical Report", "track": "post-training", "matched_queries": [ "reward-model" ], "source_provenance": [] }, { "id": "2609.02931", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "LLM-Guided Reinforcement Learning for Adaptive NPC Behavior in Multi-Agent Combat Games", "track": "post-training", "matched_queries": [ "llm-rl" ], "source_provenance": [] } ] }, { "batch": "recent-20260907-agent", "date": "2026-09-07", "scope_kind": "automated-review-closure", "required_tracks": [ "agent" ], "source_artifact": "/tmp/scan-20260907-agent.json", "window": { "start": "2026-09-03", "end": "2026-09-07" }, "candidates": [ { "id": "2609.05395", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Multi-Step Tool-Calling over Korean Open Public APIs: A Benchmark and a Data-Synthesis Recipe", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.05385", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Necessary or Sufficient? Evaluating LLM Explanations With Behavioural Evidence", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.05374", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "CUA-Universe: A Scalable and Dynamic Environment for Hybrid GUI+CLI Agents", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.05339", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Does Your Agent's Memory Survive a Model Upgrade? A Controlled Study of Memory Portability", "track": "agent", "matched_queries": [ "agent-memory" ], "source_provenance": [] }, { "id": "2609.05335", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "The History Is the Detector: Executing CVE Patch History, End-to-End", "track": "agent", "matched_queries": [ "llm-agent", "software-agent" ], "source_provenance": [] }, { "id": "2609.05324", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "RoboSPA: Can VLA Models Go Beyond Simple Scenes and Short-Horizon Tasks?", "track": "agent", "matched_queries": [ "agent-memory", "agent-planning" ], "source_provenance": [] }, { "id": "2609.05314", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Large Language Models for HVAC Operations in Building Energy Systems: A Critical Review of Methods, Applications, and Deployment Readiness", "track": "agent", "matched_queries": [ "agentic-rl", "llm-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2609.05298", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Online Change-point Detection for Cooperative Multi-Agent Reinforcement Learning", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.05279", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Testing Interchangeability in LLM Agent Teams", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.05269", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "CONTINUITY: Security-Context Contracts for Composable LLM Agent Controls", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.05261", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Trace2Tower: Transition-Aware EigenTrace Induction of Multi-Level Skills for LLM Agents", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent", "software-agent" ], "source_provenance": [] }, { "id": "2609.05258", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Ask Before You Optimize: Dynamic Pre-Formulation Clarification for Interactive Optimization", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.05227", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "CABAL: Multi-Agent Simulacra for Tracing the Effects of Collusive Bidding in Peer Review", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.05133", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "A Schema Bounded Language Model for Refining Robot Policies Without Destabilizing Local Learning", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.05079", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "TruthInsightBench: An Evidence-Grounded Benchmark for Automated Evaluation of Open-Ended Scientific Discovery Agents", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.05069", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "A Structured Debate-Mixture-of-Agents Framework for Complex Clinical Diagnostic Decision Support", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.05037", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "How do LLMs Evaluate Perceived Moral Agency? Investigating Moral Decision-Making in Human-Artificial Agents Interactions", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.05036", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Moral Competence Before Moral Content: Why LLM Agents Lack the Prerequisites for Coherent Alignment", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.05019", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "TROVE: Adaptive Agent Skill Orchestration via Trace-Grounded Route Validation and Editing", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.05018", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "How a Chatbot's Response Style Shapes a Classroom: A Multi-Agent Simulation of Students Consulting AI", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04999", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "BIT.UA at BioASQ 14B: Modular Retrieval with pg_textsearch and Qdrant, and Agent-Based Answer Generation", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04917", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Artificial Intelligence in Equity and Crypto Markets: Progress, Profitability Evidence, and the Limits of Automated Investing", "track": "agent", "matched_queries": [ "agentic-rl", "tool-agent" ], "source_provenance": [] }, { "id": "2609.04915", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Compact-Memory LLM Agents via Online Max-Member Clustering and Atom-Aware Packing", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04913", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "ARIA - An Agentic Framework for Autonomous Testing of Infotainment Systems", "track": "agent", "matched_queries": [ "llm-agent", "software-agent" ], "source_provenance": [] }, { "id": "2609.04894", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "From Language Models to World-Acting Systems: Progress and Limits of Agentic AI across Digital, Social, Virtual, and Physical Environments", "track": "agent", "matched_queries": [ "agent-planning", "multi-agent" ], "source_provenance": [] }, { "id": "2609.04880", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Reinforcement Learning for Sequential Solar PV Policy Design under Uncertainty: An Agent-Based Approach", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.04875", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Forgetting Without Restarting: Execution-State Unlearning for Stateful LLM Agents", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04871", "key": "autolr", "status": "implemented", "priority": "P0", "doc": "reproductions/2609.04871-autolr/README.md", "title": "AutoLR: Automating the Path from Research to Launch Review in Industrial Recommender Systems", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.04866", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "LLM-Assisted Behavioural and Scenario Augmentation for Agent-Based Energy Adoption Models", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04865", "key": "coskill", "status": "implemented", "priority": "P0", "doc": "agent-research/2609.04865-coskill/README.md", "title": "CoSkill: Joint Reinforcement Learning of Reasoning and Meta-Skill Agents for Hierarchical Skill Evolution", "track": "agent", "matched_queries": [ "agentic-rl", "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.04841", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "MABPD: Multi-Agent Bias Probing & Detection via Structured Argument Debate", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.04820", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Cost-Aware Hierarchical Multi-Agent Ransomware Detection and Family Attribution", "track": "agent", "matched_queries": [ "agent-memory", "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.04778", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Diffusion Language Models for Mobile Edge Agentic AI: Foundations, Applications, and Challenges", "track": "agent", "matched_queries": [ "agent-memory", "llm-agent" ], "source_provenance": [] }, { "id": "2609.04773", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Persistent Teacher Anchoring for Tool-Using Agents", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04749", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "DCFA: Dual-view Causal-inspired Attribution for Failure Reasoning in LLM-based Multi-agent Systems", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.04748", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Same Request, Different Answer: Quantization Amplifies Cache-Induced Divergence in LLM Serving", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.04711", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Building a research-software catalog with a coding agent: from hackathon prototype to public deployment", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.04699", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Model Retirement Creates Reproducibility Risk in Biomedical AI Publications", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04697", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "SQL-Zero: Self-Evolving Text-to-SQL", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04681", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Beyond Code Generation: Reliability, Verification, and Cost Economics in the Agentic Software Development Lifecycle", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.04680", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "How Developers Discuss Generative AI: A Longitudinal Study of the Visual Studio Code Community", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.04667", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "ERPBench: Evaluating LLM Agents for Enterprise Decision-Making Across Competitive Market Ecologies", "track": "agent", "matched_queries": [ "agent-planning", "llm-agent" ], "source_provenance": [] }, { "id": "2609.04630", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Software Engineering in the Agent Era From Trustworthy Change to Human Agent Software Organizations", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.04629", "key": "silr", "status": "implemented", "priority": "P1", "doc": "agent-research/2609.04629-silr/README.md", "title": "SiLR: Structure-Preserving Admission and Process Reward for LLM Tool Agents", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04611", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "$τ^τ$-Bench: An Environment for End-To-End, Realistic Agent Construction", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04570", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Dynamic Adaptation of the LLM Context for Generating Routines with Coupled Semantics", "track": "agent", "matched_queries": [ "llm-agent", "software-agent" ], "source_provenance": [] }, { "id": "2609.04564", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "La Agente Óptima: Towards Agentic Self-Driving Laboratories", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.04535", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "An Empirical Analysis of CodeQL False Positives and Query Refinements for Java Vulnerabilities", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.04523", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "MaxKernel: Agentic Kernel Generation for TPUs", "track": "agent", "matched_queries": [ "agent-planning", "llm-agent", "multi-agent" ], "source_provenance": [] }, { "id": "2609.04518", "key": "multi-harness-rl", "status": "implemented", "priority": "P1", "doc": "agent-research/2609.04518-multi-harness-rl/README.md", "title": "What Does Multi-Harness RL Learn? Credit Assignment and Portability in Coding Agents", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.04445", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Conformity Breaks Conformal Prediction", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04444", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "HarvestBench: Measuring Whether LLM Agents Will Pay to Avoid Killing Animals", "track": "agent", "matched_queries": [ "agentic-rl", "llm-agent" ], "source_provenance": [] }, { "id": "2609.04442", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "GRACE: Graph-Grounded Reflective Agent Copilot Engine for Expert-in-the-Loop Knowledge Expansion", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04373", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Why Better Models Can Create Riskier Systems: Evidence from LLM Agents in Financial Markets", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04366", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "VERGE: Verification-Enhanced Refinement for Grounded Extraction of Early-Onset Colorectal Cancer Symptoms in Clinical Notes", "track": "agent", "matched_queries": [ "tool-agent" ], "source_provenance": [] }, { "id": "2609.04336", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "MedProb: Probing Internal Representations of Vision-Language Models for Medical Question Answering", "track": "agent", "matched_queries": [ "multi-agent" ], "source_provenance": [] }, { "id": "2609.04303", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Abstraction Agent", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04290", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Evidence Integration in Large Language Models", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04286", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "From Matching Models to Recruiting Agents: A Systematized Narrative Review of AI Recruitment Systems, Evaluation, and Governance", "track": "agent", "matched_queries": [ "llm-agent", "multi-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2609.04280", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "EVOHARNESSBENCH: Can Your Agents Keep Pace with an Evolving Harness?", "track": "agent", "matched_queries": [ "llm-agent" ], "source_provenance": [] }, { "id": "2609.04270", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Reviewer Capability Governs Rejection Targeting, Not Repair Skill: Evidence from LLM Execute-Review-Revise Pipelines", "track": "agent", "matched_queries": [ "llm-agent", "software-agent" ], "source_provenance": [] }, { "id": "2609.00070", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "AutoXRD: Autonomous LLM Agents and Comprehensive Evaluation for Powder Diffraction Analysis", "track": "agent", "matched_queries": [ "agent-planning" ], "source_provenance": [] }, { "id": "2609.00058", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "CUDA-Harness: Harnessing Agentic CUDA Kernel Generation and Optimization from Natural Language", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.00052", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "AgentProv: Auditing Agentic LLM API Providers via Tool-use Policy Probes", "track": "agent", "matched_queries": [ "tool-agent" ], "source_provenance": [] }, { "id": "2609.00050", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Towards Agentic Cloud Engineering: Graph and Loop Engineering with a Zero-Trust Agent Harness", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.00038", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "trajectory-judge: What Outcome-Only LLM Judges Miss on Agent Trajectories", "track": "agent", "matched_queries": [ "software-agent", "tool-agent" ], "source_provenance": [] }, { "id": "2609.00035", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "SilentProbe: Measuring Silent Failure in Production APIs Used as Agent Tools", "track": "agent", "matched_queries": [ "software-agent" ], "source_provenance": [] }, { "id": "2609.01641", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "SocialBuddy: Tailoring Search Agent for Social Scenarios", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.00028", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "UI-Venus-2 Technical Report", "track": "agent", "matched_queries": [ "web-agent" ], "source_provenance": [] }, { "id": "2609.02931", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "LLM-Guided Reinforcement Learning for Adaptive NPC Behavior in Multi-Agent Combat Games", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.02928", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "Towards Scaling Reinforcement Learning to Massive Populations: Learning Mean-Field Representations", "track": "agent", "matched_queries": [ "agentic-rl" ], "source_provenance": [] }, { "id": "2609.00015", "status": "rejected", "priority": "P2", "reason": "高召回候选经摘要、论文正文入口与可执行性复核：未满足工业量化线上证据门槛，或相对现有实现缺少足够独立且可公平验证的核心机制。", "title": "OpenAgentFlow: Enabling System-Wide Safety Boundaries for Heterogeneous AI Agent Fleets", "track": "agent", "matched_queries": [ "agent-planning" ], "source_provenance": [] } ] }, { "batch": "recent-20260912-cross-domain", "scope_kind": "partial-cross-source-review", "checked_at": "2026-09-12T23:00:00+08:00", "window": "2026-09-05 through 2026-09-12", "sources": [ "DeepXiv candidate search", "arXiv abstract pages", "arXiv HTML and PDF full text", "official repositories" ], "required_tracks": [ "recommendation", "foundation-model", "post-training", "agent" ], "coverage": { "arxiv_api": "rate-limited (HTTP 429); primary full text and DeepXiv fallback used", "coverage_complete": false, "watermark_advanced": false, "overlap_start": "2026-09-05", "source_artifact": "recent-paper-scan-20260912.md" }, "candidates": [ {"id": "2609.11052", "priority": "P0", "track": "recommendation", "status": "implemented", "key": "unirec", "doc": "reproductions/2609.11052-unirec/README.md"}, {"id": "2609.08443", "priority": "P0", "track": "recommendation", "status": "implemented", "key": "sequenceo1", "doc": "reproductions/2609.08443-sequenceo1/README.md"}, {"id": "2609.08725", "priority": "P0", "track": "recommendation", "status": "implemented", "key": "baff", "doc": "reproductions/2609.08725-baff/README.md"}, {"id": "2609.08248", "priority": "P1", "track": "foundation-model", "status": "implemented", "key": "research-portfolio-amle", "doc": "evolution/2609.08248-amle/README.md"}, {"id": "2609.10922", "priority": "P1", "track": "foundation-model", "status": "implemented", "key": "research-portfolio-auto-recsys", "doc": "evolution/2609.10922-auto-recsys/README.md"}, {"id": "2609.08798", "priority": "P1", "track": "post-training", "status": "implemented", "key": "oprd", "doc": "post-training/2609.08798-oprd/README.md"}, {"id": "2609.08337", "priority": "P1", "track": "post-training", "status": "implemented", "key": "route-opd", "doc": "post-training/2609.08337-route-opd/README.md"}, {"id": "2609.10154", "priority": "P1", "track": "post-training", "status": "implemented", "key": "compass-opd", "doc": "post-training/2609.10154-compass-opd/README.md"}, {"id": "2609.09135", "priority": "P1", "track": "post-training", "status": "implemented", "key": "probe-erpo", "doc": "post-training/2609.09135-probe-erpo/README.md"}, {"id": "2609.09153", "priority": "P1", "track": "agent", "status": "implemented", "key": "procedural-graphs", "doc": "agent-research/2609.09153-procedural-graphs/README.md"}, {"id": "2609.08273", "priority": "P1", "track": "agent", "status": "implemented", "key": "memforest", "doc": "agent-research/2609.08273-memforest/README.md"}, {"id": "2609.08404", "priority": "P1", "track": "agent", "status": "implemented", "key": "feedback-scaffold", "doc": "agent-research/2609.08404-feedback-scaffold/README.md"}, {"id": "2609.11636", "priority": "P1", "track": "agent", "status": "implemented", "key": "maple", "doc": "agent-research/2609.11636-maple/README.md"} ] }, { "batch": "recent-20260914-cross-domain", "scope_kind": "reviewed-primary-source-batch", "checked_at": "2026-09-14T00:45:00+08:00", "window": "2026-09-09 through 2026-09-11", "sources": [ "arXiv recent pages, HTML and PDF full text", "Google Research official publications and blog", "official author repositories" ], "required_tracks": [ "recommendation", "foundation-model", "post-training", "agent" ], "coverage": { "coverage_complete": true, "watermark_advanced": true, "watermark": "2026-09-11", "overlap_start": "2026-09-09", "source_artifact": "recent-paper-scan-20260914.md" }, "candidates": [ {"id": "2609.11752", "priority": "P0", "track": "recommendation", "status": "implemented", "key": "sirf", "doc": "reproductions/2609.11752-sirf/README.md"}, {"id": "2609.11699", "priority": "P0", "track": "post-training", "status": "implemented", "key": "nsd", "doc": "post-training/2609.11699-nsd/README.md"}, {"id": "2609.11682", "priority": "P0", "track": "agent", "status": "implemented", "key": "cobra-skills", "doc": "agent-research/2609.11682-cobra-skills/README.md"}, {"id": "2609.11677", "priority": "P0", "track": "agent", "status": "implemented", "key": "ecdysis", "doc": "agent-research/2609.11677-ecdysis/README.md"}, {"id": "2609.11060", "priority": "P0", "track": "agent", "status": "implemented", "key": "grounded-memory", "doc": "agent-research/2609.11060-grounded-memory/README.md"}, {"id": "2609.11582", "priority": "P0", "track": "foundation-model", "status": "implemented", "key": "omnikvquant", "doc": "foundation-models/2609.11582-omnikvquant/README.md"}, {"id": "2508.04086", "priority": "P0", "track": "agent", "status": "implemented", "key": "toolgrad", "doc": "agent-research/2508.04086-toolgrad/README.md"}, {"id": "mlsys2026-prompts", "priority": "P0", "track": "agent", "status": "implemented", "key": "prompts", "doc": "agent-research/mlsys2026-prompts/README.md"}, {"id": "2609.11929", "priority": "P1", "track": "foundation-model", "status": "implemented", "key": "sensenova-u1-5", "doc": "foundation-models/2609.11929-sensenova-u1-5/README.md"}, {"id": "2609.11899", "priority": "P1", "track": "foundation-model", "status": "implemented", "key": "frames-on-demand", "doc": "foundation-models/2609.11899-frames-on-demand/README.md"}, {"id": "2609.11917", "priority": "P1", "track": "foundation-model", "status": "implemented", "key": "repeat-aware-moe", "doc": "foundation-models/2609.11917-repeat-aware-moe/README.md"}, {"id": "2609.11768", "priority": "P1", "track": "post-training", "status": "implemented", "key": "adaptive-opd-gate", "doc": "post-training/2609.11768-adaptive-opd-gate/README.md"}, {"id": "2609.11739", "priority": "P1", "track": "post-training", "status": "implemented", "key": "locus", "doc": "post-training/2609.11739-locus/README.md"}, {"id": "2609.11655", "priority": "P1", "track": "foundation-model", "status": "implemented", "key": "musec", "doc": "foundation-models/2609.11655-musec/README.md"}, {"id": "2609.11414", "priority": "P1", "track": "foundation-model", "status": "implemented", "key": "swrouter", "doc": "foundation-models/2609.11414-swrouter/README.md"}, {"id": "2609.11393", "priority": "P1", "track": "post-training", "status": "implemented", "key": "tasco", "doc": "post-training/2609.11393-tasco/README.md"}, {"id": "2609.10901", "priority": "P1", "track": "agent", "status": "implemented", "key": "searchatlas", "doc": "agent-research/2609.10901-searchatlas/README.md"}, {"id": "2609.10750", "priority": "P1", "track": "agent", "status": "implemented", "key": "skill-retention", "doc": "agent-research/2609.10750-skill-retention/README.md"}, {"id": "2609.11042", "priority": "P1", "track": "agent", "status": "implemented", "key": "t1-terminal-rl", "doc": "agent-research/2609.11042-t1-terminal-rl/README.md"} ] }, { "batch": "recent-20260915-cross-domain", "scope_kind": "partial-primary-source-review", "checked_at": "2026-09-15T12:00:00+08:00", "window": "2026-09-10 through 2026-09-15", "sources": [ "arXiv official recent pages, abstracts, HTML and PDF full text", "official author repositories", "full-text production experiment sections" ], "required_tracks": [ "recommendation", "foundation-model", "post-training", "agent" ], "coverage": { "arxiv_api": "connection reset; official recent pages used as fallback", "coverage_complete": false, "watermark_advanced": false, "watermark": "2026-09-11", "overlap_start": "2026-09-10", "source_artifact": "recent-paper-scan-20260915.md" }, "candidates": [ { "id": "2609.11943", "priority": "P0", "track": "recommendation", "status": "implemented", "key": "pindco", "doc": "reproductions/2609.11943-pindco/README.md" }, { "id": "2609.12842", "priority": "P0", "track": "recommendation", "status": "implemented", "key": "mima", "doc": "reproductions/2609.12842-mima/README.md" }, { "id": "2609.12375", "priority": "P0", "track": "recommendation", "status": "implemented", "key": "chronicle-rec", "doc": "reproductions/2609.12375-chronicle-rec/README.md" }, { "id": "2609.13141", "priority": "P1", "track": "foundation-model", "status": "implemented", "key": "sas-attention", "doc": "reproductions/2609.13141-sas-attention/README.md" }, { "id": "2609.12399", "priority": "P1", "track": "recommendation", "status": "deferred", "reason": "Defining contribution is a fused GPU large-beam linear-attention decode path; requires real A100/A30 correctness, memory and latency validation." }, { "id": "2609.13060", "priority": "P1", "track": "post-training", "status": "deferred", "reason": "Requires a public diffusion LM, teacher-canvas curriculum and a reproducible RL/tool-use training budget." }, { "id": "2609.12191", "priority": "P1", "track": "agent", "status": "deferred", "reason": "Requires released transcripts and blind human labels to reproduce grounded-reward ranking and construct validity." }, { "id": "2609.12471", "priority": "P1", "track": "agent", "status": "deferred", "reason": "Defining HIP/ROCm execution environment cannot be validated on the available NVIDIA A100/A30 hosts." } ] } ] }