{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.10479", "kind": "paper", "url": "https://arxiv.org/abs/2606.10479"}, {"id": "gh:simplified-reasoning/combench", "kind": "repo", "url": "https://github.com/Simplified-Reasoning/ComBench"}, {"kind": "website", "url": "https://simplified-reasoning.github.io/ComBench/docs/"}], "authors": ["Shunkai Zhang", "Haoran Zhang", "Yun Luo", "Qianjia Cheng", "Haodi Lei", "Yizhuo Li", "Runzhe Zhan", "Zhilin Wang", "Bangjie Xu", "Yucheng Su", "Xinmiao Han", "Xiaoye Qu", "Dongrui Liu", "Zhouchen Lin", "Yu Qiao", "Ning Ding", "Yafu Li", "Yu Cheng"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.10479", "source_url": "https://arxiv.org/abs/2606.10479"}, "categories": ["Language & Knowledge", "Mathematics & Formal Science", "Reasoning"], "description": {"en": "ComBench evaluates large language models on 100 human-annotated Olympiad-level combinatorics problems, split into 50 analysis-centric and 50 construction-centric tasks. Scoring combines rubric-guided proof grading with deterministic verification of construction outputs."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.10479", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.10479", "title": "ComBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Shanghai Artificial Intelligence Laboratory", "Chinese University of Hong Kong", "Tsinghua University", "Peking University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.10479", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.10479", "source_url": "https://openalex.org/W7164204358"}, "key": "claire-radar:2606.10479", "modality": null, "name": "ComBench", "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_combench_b27be297", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "7e3a65324a7284095dc0547e48207b4ac06a36a047bfb284ce133ab8b0ac1517", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.10479"}, "publisher": null, "released": "2026-06-09", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.10479", "source_url": "https://arxiv.org/abs/2606.10479"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-10479", "source": "claire_radar", "source_benchmark_id": "2606.10479", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["Mathematics & Formal Science"], "area": "Language & Knowledge", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/Simplified-Reasoning/ComBench", "githubScope": "benchmark_repo", "githubStars": 13, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 21, "hfPaperUrl": "https://huggingface.co/papers/2606.10479", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "lastSuccessfulDate": "2026-09-24", "state": "stale"}}, "status": "fresh"}, "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": "b0e3f91093bc15da60e71ed07dcd7dcac15cba99db7a2fa710aecffbd7e374ff", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "ComBench evaluates large language models on 100 human-annotated Olympiad-level combinatorics problems, split into 50 analysis-centric and 50 construction-centric tasks. Scoring combines rubric-guided proof grading with deterministic verification of construction outputs.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "We introduce ComBench, an Olympiad-level combinatorics benchmark for evaluating and diagnosing the combinatorial reasoning capabilities of large language models."}, "familyId": "bmf_29e41fd248d6", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [{"evidence": [{"basis": "existing-code-link", "linkedUrl": "https://github.com/Simplified-Reasoning/ComBench", "url": "https://github.com/Simplified-Reasoning/ComBench"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/Simplified-Reasoning/ComBench", "url": "https://huggingface.co/papers/2606.10479"}, {"basis": "huggingface-paper-association", "linkedUrl": "https://github.com/Simplified-Reasoning/ComBench", "url": "https://huggingface.co/api/papers/2606.10479"}, {"basis": "linked-from-project", "linkedUrl": "https://github.com/Simplified-Reasoning/ComBench", "url": "https://simplified-reasoning.github.io/ComBench/docs/"}], "scope": "benchmark_repo", "url": "https://github.com/Simplified-Reasoning/ComBench", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_combench_b27be297", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/Simplified-Reasoning/ComBench", "data": null, "hfPaper": "https://huggingface.co/papers/2606.10479", "paper": "https://arxiv.org/abs/2606.10479", "pdf": "https://arxiv.org/pdf/2606.10479", "project": "https://simplified-reasoning.github.io/ComBench/docs/", "report": "https://arxiv.org/abs/2606.10479"}, "metrics": [], "motivation": "Combinatorics is central to Olympiad-level mathematical problem solving, requiring deep discrete reasoning, creative constructions, and rigorous structural insight.", "name": "ComBench", "oneLine": "ComBench evaluates large language models on 100 human-annotated Olympiad-level combinatorics problems, split into 50 analysis-centric and 50 construction-centric tasks. Scoring combines rubric-guided proof grading with deterministic verification of construction outputs.", "paperTitle": "ComBench: A Benchmark for Rigorous Proof Reasoning and Constructive Realization in Olympiad-Level Combinatorics", "primaryDomain": "Mathematics & Formal Science", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.10479", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "Simplified Reasoning", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/Simplified-Reasoning/ComBench"}], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-09", "source": {"authors": ["Shunkai Zhang", "Haoran Zhang", "Yun Luo", "Qianjia Cheng", "Haodi Lei", "Yizhuo Li", "Runzhe Zhan", "Zhilin Wang", "Bangjie Xu", "Yucheng Su", "Xinmiao Han", "Xiaoye Qu", "Dongrui Liu", "Zhouchen Lin", "Yu Qiao", "Ning Ding", "Yafu Li", "Yu Cheng"], "categories": ["cs.AI"], "id": "2606.10479", "title": "ComBench: A Benchmark for Rigorous Proof Reasoning and Constructive Realization in Olympiad-Level Combinatorics", "type": "arxiv", "url": "https://arxiv.org/abs/2606.10479"}, "sourceUpdatedAt": "2026-06-09T06:50:15Z", "topics": ["Reasoning"], "venueAttempts": [], "whyItMatters": "ComBench fills a gap in evaluating creative and rigorous combinatorial reasoning at the Olympiad level. It provides separate scores for proof quality and construction validity, helping diagnose where models diverge in these capabilities, and enables fine-grained comparison of frontier models."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
