{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2607.26397", "kind": "paper", "url": "https://arxiv.org/abs/2607.26397"}], "authors": ["Linyu Li", "Zhi Jin", "Yichi Zhang", "Dongming Jin", "Yuanpeng He", "Huanyao Zhang", "Xuan Zhang", "Gadeng Luosang", "Nyima Tashi"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.26397", "source_url": "https://arxiv.org/abs/2607.26397"}, "categories": ["Biology & Drug Discovery", "Factuality", "Language & Knowledge", "Reasoning"], "description": {"en": "EC-Reason-Bench is a training-free diagnostic protocol for LLM enzyme classification, isolating four reasoning levers: output structure, external knowledge, reasoning structure, and robustness, to measure why LLMs fail on EC number prediction."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.26397", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.26397", "title": "EC-Reason-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Peking University", "Zhejiang University", "Tibet University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2607.26397", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2607.26397", "source_url": "https://openalex.org/W7171795843"}, "key": "claire-radar:2607.26397", "modality": null, "name": "EC-Reason-Bench", "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_ec-reason-bench_6994eec4", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "680c0f0e384af7ee4325c02650a21cbe7b658502a9ae40034addefdf903da919", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2607.26397"}, "publisher": null, "released": "2026-07-29", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.26397", "source_url": "https://arxiv.org/abs/2607.26397"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-26397", "source": "claire_radar", "source_benchmark_id": "2607.26397", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Mixed", "applicationDomains": ["Biology & Drug Discovery"], "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": null, "hfPaperUrl": "https://huggingface.co/papers/2607.26397", "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", "state": "unavailable"}}, "status": "unavailable"}, "capabilities": ["Reasoning", "Factuality"], "confidence": "High", "construction": "Aggregate Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "60aac0fe5509c09669f763909ee7d0f84bd843e63f84d7e5e3df9a9f64038552", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "EC-Reason-Bench is a training-free diagnostic protocol for LLM enzyme classification, isolating four reasoning levers: output structure, external knowledge, reasoning structure, and robustness, to measure why LLMs fail on EC number prediction.", "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 propose EC-Reason-Bench, a training-free, diagnostic evaluation protocol built to answer two questions: why general LLMs score close to nothing on EC number prediction, and how much of that loss can be recovered without updating a single weight."}, "familyId": "bmf_e8fbcd70b60f", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_ec-reason-bench_6994eec4", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Pharma & Biotech"], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2607.26397", "paper": "https://arxiv.org/abs/2607.26397", "pdf": "https://arxiv.org/pdf/2607.26397", "project": null, "report": "https://arxiv.org/abs/2607.26397"}, "metrics": [], "motivation": "Enzyme function prediction is a hierarchical, knowledge-intensive form of protein function classification.", "name": "EC-Reason-Bench", "oneLine": "EC-Reason-Bench is a training-free diagnostic protocol for LLM enzyme classification, isolating four reasoning levers: output structure, external knowledge, reasoning structure, and robustness, to measure why LLMs fail on EC number prediction.", "paperTitle": "Knowledge before Reasoning: EC-Reason-Bench, a Training-Free Diagnostic Benchmark for LLM Enzyme Classification", "primaryDomain": "Biology & Drug Discovery", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.26397", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {"90d": {"confidence": "Low", "coverage": 0.0, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 90}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": null, "value": null}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "rank": null, "score": null}, "method": "attention-ranking-v13", "rank": null, "score": null}}, "readiness": "Paper only", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-07-29", "source": {"authors": ["Linyu Li", "Zhi Jin", "Yichi Zhang", "Dongming Jin", "Yuanpeng He", "Huanyao Zhang", "Xuan Zhang", "Gadeng Luosang", "and Nyima Tashi"], "categories": ["cs.CL", "cs.LG", "q-bio.QM"], "id": "2607.26397", "title": "Knowledge before Reasoning: EC-Reason-Bench, a Training-Free Diagnostic Benchmark for LLM Enzyme Classification", "type": "arxiv", "url": "https://arxiv.org/abs/2607.26397"}, "sourceUpdatedAt": "2026-07-29T02:16:09Z", "topics": ["Reasoning"], "venueAttempts": [], "whyItMatters": "The protocol reveals that external knowledge is decisive and that reasoning acts as an arbiter among conflicting neighbors, showing that single-number leaderboards obscure the source of accuracy loss. This informs how to design knowledge-integration methods for protein function prediction."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
