{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2609.00787", "kind": "paper", "url": "https://arxiv.org/abs/2609.00787"}, {"id": "gh:thunlp/studybench", "kind": "repo", "url": "https://github.com/thunlp/StudyBench"}], "authors": ["Yinghao Chen", "Zixi Chen", "Bingxiang He", "Ziqing Qiao", "Huan-ang Gao", "Yinuo Xu", "Yuxin Zuo", "Zeyuan Liu", "Yuhao Zhan", "Chaojun Xiao"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2609.00787", "source_url": "https://arxiv.org/abs/2609.00787"}, "categories": ["Language & Knowledge", "Mathematics & Formal Science", "cs.AI"], "description": {"en": "Evaluates language models on hard physics textbook exercises (Application Set) and olympiad-level theory problems (Transfer Set), reporting pass@k accuracy for absorption and transfer of textbook knowledge."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2609.00787", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2609.00787", "title": "StudyBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2609.00787", "modality": null, "name": "StudyBench", "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": "true", "origin_record_id": "bm_studybench_38f4772e", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "af61f2ec50b855c770d5214b3f7992cdc3504c99587ef87a77bf7802f2211306", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-09-03T00:59:34.317835Z", "source_url": "https://arxiv.org/abs/2609.00787"}, "publisher": null, "released": "2026-09-01", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2609.00787", "source_url": "https://arxiv.org/abs/2609.00787"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2609-00787", "source": "claire_radar", "source_benchmark_id": "2609.00787", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["Mathematics & Formal Science"], "area": "Language & Knowledge", "attention": {"asOf": "2026-09-02", "attemptedAt": "2026-09-02T05:12:51.621571Z", "githubRepo": "https://github.com/thunlp/StudyBench", "githubScope": "benchmark_repo", "githubStars": 1, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 1, "hfPaperUrl": "https://huggingface.co/papers/2609.00787", "observedAt": "2026-09-02T05:12:51.621571Z", "signalStatus": {"githubStars": {"observedAt": "2026-09-02T05:12:51.621571Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"observedAt": "2026-09-02T05:12:51.621571Z", "state": "fresh"}}, "status": "fresh"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "The niche olympiad physics domain and self-evolution focus limit immediate reach, but acceptance at EMNLP 2026 Findings and reusable assets may attract a modest early audience.", "score": 38}, "capabilities": [], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-03T00:59:34.317835Z", "inputHash": "0848aed228164d1cb648590e39366893d968ca8f7b2defc4c2eeccee4ea64c68", "model": "deepseek-v4-pro"}, "curation": {"canonicalNameEvidence": "We introduce StudyBench, a controlled physics benchmark that directly measures how efficiently a self-evolution method converts training material into capability.", "canonicalNameSource": "abstract", "decisionReason": "The benchmark has a formally named release, a clear scoring protocol with described splits and metrics, and publicly available code and test sets in the repository.", "model": "deepseek-v4-pro", "reviewedAt": "2026-09-03T00:59:34.317835Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Evaluates language models on hard physics textbook exercises (Application Set) and olympiad-level theory problems (Transfer Set), reporting pass@k accuracy for absorption and transfer of textbook knowledge.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "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 StudyBench, a controlled physics benchmark that directly measures how efficiently a self-evolution method converts training material into capability."}, "familyId": "bmf_fad9b75d6b9d", "firstSeenAt": "2026-09-02", "heat": null, "id": "bm_studybench_38f4772e", "indexedAt": "2026-09-02T05:21:06.631990Z", "industrySectors": [], "links": {"code": "https://github.com/thunlp/StudyBench", "data": null, "paper": "https://arxiv.org/abs/2609.00787", "pdf": "https://arxiv.org/pdf/2609.00787", "project": null, "report": "https://arxiv.org/abs/2609.00787"}, "metrics": [], "motivation": "Humans need to study only a handful of well-written textbooks to master a discipline and attempt its hardest problems.", "name": "StudyBench", "oneLine": "Evaluates language models on hard physics textbook exercises (Application Set) and olympiad-level theory problems (Transfer Set), reporting pass@k accuracy for absorption and transfer of textbook knowledge.", "paperTitle": "StudyBench: Can Self-Evolution Squeeze Textbooks for Olympiad Capability?", "primaryDomain": "Mathematics & Formal Science", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2609.00787", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-09-02T05:21:06.631990Z"}, "publications": [], "publishers": [{"name": "THUNLP", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/thunlp/StudyBench"}], "ranking": {"30d": {"confidence": "High", "coverage": 0.85, "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": 30}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": 0.10323645750017778, "value": 1}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": 0.1228568942814387, "value": 1}}, "confidence": "High", "coverage": 0.85, "rank": 188, "score": 17}, "method": "attention-ranking-v13", "rank": 188, "score": 17}, "90d": {"confidence": "Medium", "coverage": 0.7, "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": 0.09434039834931938, "value": 1}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": 0.1228568942814387, "value": 1}}, "confidence": "Medium", "coverage": 0.7, "rank": 393, "score": 22}, "method": "attention-ranking-v13", "rank": 393, "score": 22}, "today": {"confidence": "Medium", "coverage": 0.7, "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": 2, "historyDate": "2026-08-31", "rank": null, "score": null, "windowDays": 1}, "level": {"components": {"githubStars": {"normalization": "percentile", "normalized": 0.5, "percentile": 0.5, "value": 1}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": 0.5, "percentile": 0.5, "value": 1}, "llmAttentionForecast": {"maxContribution": 0.25, "percentile": 0.0, "role": "bonus", "value": null}}, "confidence": "Medium", "coverage": 0.7, "rank": 1, "score": 38}, "method": "attention-ranking-v13", "rank": 1, "score": 38}}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-09-01", "source": {"authors": ["Yinghao Chen", "Zixi Chen", "Bingxiang He", "Ziqing Qiao", "Huan-ang Gao", "Yinuo Xu", "Yuxin Zuo", "Zeyuan Liu", "Yuhao Zhan", "Chaojun Xiao"], "categories": ["cs.AI"], "id": "2609.00787", "title": "StudyBench: Can Self-Evolution Squeeze Textbooks for Olympiad Capability?", "type": "arxiv", "url": "https://arxiv.org/abs/2609.00787"}, "sourceUpdatedAt": "2026-09-01T06:33:53Z", "topics": ["cs.AI"], "venueAttempts": [], "whyItMatters": "Provides a controlled setup to separate a method's ability to absorb training material from its ability to transfer knowledge to harder problems, addressing a gap in evaluating self-evolution efficiency."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
