{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:OHRCXXQJNCX7MO4JN2OAV3RPDI","short_pith_number":"pith:OHRCXXQJ","schema_version":"1.0","canonical_sha256":"71e22bde0968aff63b896e9c0aee2f1a0f536f45943012c306b329088e26addf","source":{"kind":"arxiv","id":"2502.16268","version":1},"attestation_state":"computed","paper":{"title":"ThinkBench: Dynamic Out-of-Distribution Evaluation for Robust LLM Reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jun Wang, Kun Shao, Leyang Cui, Linyi Yang, Qingcheng Zeng, Shuang Chen, Shulin Huang, Weinan Zhang, Yan Song, Ying Wen, Yue Zhang, Ziyu Wan","submitted_at":"2025-02-22T15:41:51Z","abstract_excerpt":"Evaluating large language models (LLMs) poses significant challenges, particularly due to issues of data contamination and the leakage of correct answers. To address these challenges, we introduce ThinkBench, a novel evaluation framework designed to evaluate LLMs' reasoning capability robustly. ThinkBench proposes a dynamic data generation method for constructing out-of-distribution (OOD) datasets and offers an OOD dataset that contains 2,912 samples drawn from reasoning tasks. ThinkBench unifies the evaluation of reasoning models and non-reasoning models. We evaluate 16 LLMs and 4 PRMs under "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2502.16268","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-22T15:41:51Z","cross_cats_sorted":[],"title_canon_sha256":"7cd21c6df1b48f04cd3347ae4de820a3c5a53b408f9abd4bae7bb80c353cf54b","abstract_canon_sha256":"e1a1fb1c0d21619c9cfe35754b3ca12597d6f996aa9029ec996d87a85b194034"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:18:24.006577Z","signature_b64":"QRz8JeTewpVZkPC57Xy5FteF5IkiFSIMhNpeffO/Lk08ysesoA04n8J8CbVpA4qR+Ljn5WhmwSAZ0JyZ8MjTAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"71e22bde0968aff63b896e9c0aee2f1a0f536f45943012c306b329088e26addf","last_reissued_at":"2026-07-05T10:18:24.006083Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:18:24.006083Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ThinkBench: Dynamic Out-of-Distribution Evaluation for Robust LLM Reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jun Wang, Kun Shao, Leyang Cui, Linyi Yang, Qingcheng Zeng, Shuang Chen, Shulin Huang, Weinan Zhang, Yan Song, Ying Wen, Yue Zhang, Ziyu Wan","submitted_at":"2025-02-22T15:41:51Z","abstract_excerpt":"Evaluating large language models (LLMs) poses significant challenges, particularly due to issues of data contamination and the leakage of correct answers. To address these challenges, we introduce ThinkBench, a novel evaluation framework designed to evaluate LLMs' reasoning capability robustly. ThinkBench proposes a dynamic data generation method for constructing out-of-distribution (OOD) datasets and offers an OOD dataset that contains 2,912 samples drawn from reasoning tasks. ThinkBench unifies the evaluation of reasoning models and non-reasoning models. We evaluate 16 LLMs and 4 PRMs under "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.16268","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2502.16268/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2502.16268","created_at":"2026-07-05T10:18:24.006142+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.16268v1","created_at":"2026-07-05T10:18:24.006142+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.16268","created_at":"2026-07-05T10:18:24.006142+00:00"},{"alias_kind":"pith_short_12","alias_value":"OHRCXXQJNCX7","created_at":"2026-07-05T10:18:24.006142+00:00"},{"alias_kind":"pith_short_16","alias_value":"OHRCXXQJNCX7MO4J","created_at":"2026-07-05T10:18:24.006142+00:00"},{"alias_kind":"pith_short_8","alias_value":"OHRCXXQJ","created_at":"2026-07-05T10:18:24.006142+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25450","citing_title":"The Generalization Spectrum: A Chromatographic Approach to Evaluating Learning Algorithms","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2606.25450","citing_title":"The Generalization Spectrum: A Chromatographic Approach to Evaluating Learning Algorithms","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2503.09567","citing_title":"Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Models","ref_index":296,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OHRCXXQJNCX7MO4JN2OAV3RPDI","json":"https://pith.science/pith/OHRCXXQJNCX7MO4JN2OAV3RPDI.json","graph_json":"https://pith.science/api/pith-number/OHRCXXQJNCX7MO4JN2OAV3RPDI/graph.json","events_json":"https://pith.science/api/pith-number/OHRCXXQJNCX7MO4JN2OAV3RPDI/events.json","paper":"https://pith.science/paper/OHRCXXQJ"},"agent_actions":{"view_html":"https://pith.science/pith/OHRCXXQJNCX7MO4JN2OAV3RPDI","download_json":"https://pith.science/pith/OHRCXXQJNCX7MO4JN2OAV3RPDI.json","view_paper":"https://pith.science/paper/OHRCXXQJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.16268&json=true","fetch_graph":"https://pith.science/api/pith-number/OHRCXXQJNCX7MO4JN2OAV3RPDI/graph.json","fetch_events":"https://pith.science/api/pith-number/OHRCXXQJNCX7MO4JN2OAV3RPDI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OHRCXXQJNCX7MO4JN2OAV3RPDI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OHRCXXQJNCX7MO4JN2OAV3RPDI/action/storage_attestation","attest_author":"https://pith.science/pith/OHRCXXQJNCX7MO4JN2OAV3RPDI/action/author_attestation","sign_citation":"https://pith.science/pith/OHRCXXQJNCX7MO4JN2OAV3RPDI/action/citation_signature","submit_replication":"https://pith.science/pith/OHRCXXQJNCX7MO4JN2OAV3RPDI/action/replication_record"}},"created_at":"2026-07-05T10:18:24.006142+00:00","updated_at":"2026-07-05T10:18:24.006142+00:00"}