{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.03660", "kind": "paper", "url": "https://arxiv.org/abs/2606.03660"}], "authors": ["Hongyu Guo", "Hao Li", "He Cao", "Gongbo Zhang", "Li Yuan"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.03660", "source_url": "https://arxiv.org/abs/2606.03660"}, "categories": ["Language & Knowledge", "Materials & Chemistry", "Reasoning"], "description": {"en": "ChemCoTBench-V2 evaluates chemical reasoning at the process level, with 5,620 samples across molecular understanding, editing, optimization, and reaction prediction. It checks intermediate steps using deterministic chemistry rules and reference traces."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.03660", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.03660", "title": "ChemCoTBench-V2"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Peking University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.03660", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.03660", "source_url": "https://openalex.org/W7163362892"}, "key": "claire-radar:2606.03660", "modality": null, "name": "ChemCoTBench-V2", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "High", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "unknown", "origin_record_id": "bm_chemcotbench-v2_5360db80", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "472ab5b13852d2b1cbe1d303c9994ad93eefad14b7aa2751e2e6bccb29dbf28b", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.03660"}, "publisher": null, "released": "2026-06-02", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.03660", "source_url": "https://arxiv.org/abs/2606.03660"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-03660", "source": "claire_radar", "source_benchmark_id": "2606.03660", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["Materials & Chemistry"], "area": "Language & Knowledge", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 0, "hfPaperUrl": "https://huggingface.co/papers/2606.03660", "observedAt": "2026-10-04T03:02:42.754940Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "lastSuccessfulDate": "2026-09-30", "state": "stale"}}, "status": "stale"}, "capabilities": ["Reasoning"], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "c3e8746331d8b2fba17760cb5543ca5052e315212ec565b0c995d37dce291805", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "ChemCoTBench-V2 evaluates chemical reasoning at the process level, with 5,620 samples across molecular understanding, editing, optimization, and reaction prediction. It checks intermediate steps using deterministic chemistry rules and reference traces.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact named benchmark artifact released in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "We introduce ChemCoTBench-V2, a rule-verifiable diagnostic benchmark for low-cost, auditable evaluation of structured, verifier-addressable chemical reasoning traces."}, "familyId": "bmf_43f17e5151db", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_chemcotbench-v2_5360db80", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Materials & Chemicals"], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2606.03660", "paper": "https://arxiv.org/abs/2606.03660", "pdf": "https://arxiv.org/pdf/2606.03660", "project": null, "report": "https://arxiv.org/abs/2606.03660"}, "metrics": [], "motivation": "Large language models are increasingly used as chemistry assistants, yet most chemistry benchmarks still score only final answers.", "name": "ChemCoTBench-V2", "oneLine": "ChemCoTBench-V2 evaluates chemical reasoning at the process level, with 5,620 samples across molecular understanding, editing, optimization, and reaction prediction. It checks intermediate steps using deterministic chemistry rules and reference traces.", "paperTitle": "From Answers to States: Verifiable Process-Level Evaluation of Chemical Reasoning in Large Language Models", "primaryDomain": "Materials & Chemistry", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.03660", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {}, "readiness": "Paper only", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-02", "source": {"authors": ["Hongyu Guo", "Hao Li", "He Cao", "Gongbo Zhang", "Li Yuan"], "categories": ["cs.AI"], "id": "2606.03660", "title": "From Answers to States: Verifiable Process-Level Evaluation of Chemical Reasoning in Large Language Models", "type": "arxiv", "url": "https://arxiv.org/abs/2606.03660"}, "sourceUpdatedAt": "2026-06-03T14:05:21Z", "topics": ["Reasoning"], "venueAttempts": [], "whyItMatters": "Existing chemistry benchmarks only score final answers, missing violations in reasoning. ChemCoTBench-V2 provides auditable, verifiable process-level evaluation with three separate signals, enabling fine-grained model comparison."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
