{"record": {"aliases": ["HOLMES: Evaluating Higher-Order Logical Reasoning in LLMs"], "artifacts": [{"id": "arxiv:2606.23238", "kind": "paper", "url": "https://arxiv.org/abs/2606.23238"}, {"id": "gh:wuyucheng2002/holmes", "kind": "repo", "url": "https://github.com/wuyucheng2002/HOLMES"}], "authors": ["Yucheng Wu", "Jundong Xu", "Mingzhen Ju", "Yue Yu", "Chenpeng Wang", "Haoxuan Li", "Liangming Pan"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.23238", "source_url": "https://arxiv.org/abs/2606.23238"}, "categories": ["Finance", "Language & Knowledge", "Reasoning"], "description": {"en": "HOLMES evaluates higher-order symbolic reasoning in language models using 1,379 natural-language problems paired with formalizations, answers, and verifiable reasoning traces across law and finance."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.23238", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.23238", "title": "HOLMES"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Peking University", "Jiangsu University", "Yixing People's Hospital"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.23238", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.23238", "source_url": "https://openalex.org/W7165653240"}, "key": "claire-radar:2606.23238", "modality": null, "name": "HOLMES", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"admission_policy_version": "2026-09-09.1", "confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-09-25T04:40:22Z", "data_status": "primary-source-indexed", "display_eligible": "true", "origin_record_id": "bm_holmes-evaluating-higher-order-logical-rea_dd77968a", "origin_source": "arxiv", "recognition_confidence": "0.7", "record_sha256": "e5ad00780476cd0b920b291ff5db2dfced6e27a93df2464e5de35a9702199ebd", "relation": "introduces", "review_model": "codex-cli-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T17:18:04.271287Z", "source_url": "https://arxiv.org/abs/2606.23238"}, "publisher": null, "released": "2026-06-22", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.23238", "source_url": "https://arxiv.org/abs/2606.23238"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-23238", "source": "claire_radar", "source_benchmark_id": "2606.23238", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["HOLMES: Evaluating Higher-Order Logical Reasoning in LLMs"], "annotation": "Unknown", "applicationDomains": ["Finance"], "area": "Language & Knowledge", "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "The benchmark targets a recognized reasoning limitation, supplies verifiable traces, and reports substantial headroom, although its specialized higher-order-logic focus may limit broad attention.", "score": 7}, "capabilities": ["Reasoning"], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T17:18:04.271287Z", "inputHash": "b2b28727ddc1daf4516a4b444572abd28f050188500c15b383f12b4d1f1a2ce6", "model": "codex-cli-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "We introduce HOLMES (Higher-Order Logic Meets real-world Explainable Symbolic reasoning)", "canonicalNameSource": "abstract", "decisionReason": "The source formally names the benchmark, identifies its evaluation task, reports accuracy as a comparable score, and provides a public code-and-dataset path.", "evaluationEvidence": {"reuse": "The project code and dataset are publicly available at https://github.com/wuyucheng2002/HOLMES.", "scoring": "average accuracy of only 50.64% and the best model reaching 59.54%", "task": "higher-order symbolic reasoning in LLMs"}, "model": "codex-cli-default", "reviewedAt": "2026-09-24T17:18:04.271287Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "HOLMES evaluates higher-order symbolic reasoning in language models using 1,379 natural-language problems paired with formalizations, answers, and verifiable reasoning traces across law and finance.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["benchmark term in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "We introduce HOLMES (Higher-Order Logic Meets real-world Explainable Symbolic reasoning), the first real-world benchmark for higher-order symbolic reasoning in LLMs, containing 1379 instances."}, "familyId": "bmf_20d2d09ce2d3", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_holmes-evaluating-higher-order-logical-rea_dd77968a", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": ["Financial Services"], "links": {"code": "https://github.com/wuyucheng2002/HOLMES", "data": null, "paper": "https://arxiv.org/abs/2606.23238", "pdf": "https://arxiv.org/pdf/2606.23238", "project": null, "report": "https://arxiv.org/abs/2606.23238"}, "metrics": [], "motivation": "Logical reasoning is essential for reliable AI, yet existing benchmarks are largely first-order-logic-centric, focusing on object-level deduction over fixed predicates.", "name": "HOLMES", "oneLine": "HOLMES evaluates higher-order symbolic reasoning in language models using 1,379 natural-language problems paired with formalizations, answers, and verifiable reasoning traces across law and finance.", "paperTitle": "HOLMES: Evaluating Higher-Order Logical Reasoning in LLMs", "primaryDomain": "Finance", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.23238", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "readiness": "Runnable", "recognitionConfidence": 0.7, "relation": "introduces", "releasedAt": "2026-06-22", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Yucheng Wu", "Jundong Xu", "Mingzhen Ju", "Yue Yu", "Chenpeng Wang", "Haoxuan Li", "Liangming Pan"], "categories": ["cs.AI"], "id": "2606.23238", "title": "HOLMES: Evaluating Higher-Order Logical Reasoning in LLMs", "type": "arxiv", "url": "https://arxiv.org/abs/2606.23238"}, "sourceUpdatedAt": "2026-06-23T08:36:37Z", "topics": ["Reasoning"], "venueAttempts": [], "whyItMatters": "It evaluates reasoning over rules, predicates, functions, constraints, and decision procedures beyond conventional first-order, object-level deduction."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
