{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:WTGFSDP2TC7HF3ESY7CSGJREF3","short_pith_number":"pith:WTGFSDP2","schema_version":"1.0","canonical_sha256":"b4cc590dfa98be72ec92c7c52326242ec69008d000cf27f5c8a86464c5a6e486","source":{"kind":"arxiv","id":"2501.17703","version":4},"attestation_state":"computed","paper":{"title":"Critique Fine-Tuning: Learning to Critique is More Effective than Learning to Imitate","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Wenhu Chen, Xiang Yue, Yubo Wang","submitted_at":"2025-01-29T15:20:30Z","abstract_excerpt":"Supervised Fine-Tuning (SFT) is commonly used to train language models to imitate annotated responses for given instructions. In this paper, we propose Critique Fine-Tuning (CFT), a method more effective than SFT for reasoning tasks. Instead of simply imitating correct responses, CFT trains models to critique noisy responses, inspired by human learning processes that emphasize critical thinking, deeper analysis, and nuanced understanding - traits often overlooked by standard SFT. To validate the effectiveness of CFT, we construct multiple critique datasets (e.g., WebInstruct, MetaMath, NuminaM"},"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":"2501.17703","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-29T15:20:30Z","cross_cats_sorted":[],"title_canon_sha256":"dad1a929e5c0182287d20e595e17062b2f802a52cc6ecc77c530b9df508f1b9c","abstract_canon_sha256":"e6f8a81cd7ea9d444b7509293b530c0687151bf1a66b02e450b96f26bbdaa706"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:41:20.097655Z","signature_b64":"qf2OFsu9yRlDZ6Ma2ayhbZr0IsphsIF+DNjX5hMlf69bFWUOLR25S60BfNMe/0AfCdT+N2+dPKdLHjgnnJfKCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b4cc590dfa98be72ec92c7c52326242ec69008d000cf27f5c8a86464c5a6e486","last_reissued_at":"2026-07-05T10:41:20.097062Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:41:20.097062Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Critique Fine-Tuning: Learning to Critique is More Effective than Learning to Imitate","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Wenhu Chen, Xiang Yue, Yubo Wang","submitted_at":"2025-01-29T15:20:30Z","abstract_excerpt":"Supervised Fine-Tuning (SFT) is commonly used to train language models to imitate annotated responses for given instructions. In this paper, we propose Critique Fine-Tuning (CFT), a method more effective than SFT for reasoning tasks. Instead of simply imitating correct responses, CFT trains models to critique noisy responses, inspired by human learning processes that emphasize critical thinking, deeper analysis, and nuanced understanding - traits often overlooked by standard SFT. To validate the effectiveness of CFT, we construct multiple critique datasets (e.g., WebInstruct, MetaMath, NuminaM"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.17703","kind":"arxiv","version":4},"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/2501.17703/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":"2501.17703","created_at":"2026-07-05T10:41:20.097121+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.17703v4","created_at":"2026-07-05T10:41:20.097121+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.17703","created_at":"2026-07-05T10:41:20.097121+00:00"},{"alias_kind":"pith_short_12","alias_value":"WTGFSDP2TC7H","created_at":"2026-07-05T10:41:20.097121+00:00"},{"alias_kind":"pith_short_16","alias_value":"WTGFSDP2TC7HF3ES","created_at":"2026-07-05T10:41:20.097121+00:00"},{"alias_kind":"pith_short_8","alias_value":"WTGFSDP2","created_at":"2026-07-05T10:41:20.097121+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.06065","citing_title":"SWE-Review: Closing the Loop on Issue Resolution with Agentic Code Review","ref_index":40,"is_internal_anchor":true},{"citing_arxiv_id":"2605.14636","citing_title":"Teaching Large Language Models When Not to Know: Learning Temporal Critique for Ex-Ante Reasoning","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30062","citing_title":"FakeVLM-R1: Internalizing Physical Laws via CoT for Synthetic Image Detection","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2602.07832","citing_title":"rePIRL: Learn PRM with Inverse RL for LLM Reasoning","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14539","citing_title":"Learning from Failures: Correction-Oriented Policy Optimization with Verifiable Rewards","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21718","citing_title":"Building a Precise Video Language with Human-AI Oversight","ref_index":66,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WTGFSDP2TC7HF3ESY7CSGJREF3","json":"https://pith.science/pith/WTGFSDP2TC7HF3ESY7CSGJREF3.json","graph_json":"https://pith.science/api/pith-number/WTGFSDP2TC7HF3ESY7CSGJREF3/graph.json","events_json":"https://pith.science/api/pith-number/WTGFSDP2TC7HF3ESY7CSGJREF3/events.json","paper":"https://pith.science/paper/WTGFSDP2"},"agent_actions":{"view_html":"https://pith.science/pith/WTGFSDP2TC7HF3ESY7CSGJREF3","download_json":"https://pith.science/pith/WTGFSDP2TC7HF3ESY7CSGJREF3.json","view_paper":"https://pith.science/paper/WTGFSDP2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.17703&json=true","fetch_graph":"https://pith.science/api/pith-number/WTGFSDP2TC7HF3ESY7CSGJREF3/graph.json","fetch_events":"https://pith.science/api/pith-number/WTGFSDP2TC7HF3ESY7CSGJREF3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WTGFSDP2TC7HF3ESY7CSGJREF3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WTGFSDP2TC7HF3ESY7CSGJREF3/action/storage_attestation","attest_author":"https://pith.science/pith/WTGFSDP2TC7HF3ESY7CSGJREF3/action/author_attestation","sign_citation":"https://pith.science/pith/WTGFSDP2TC7HF3ESY7CSGJREF3/action/citation_signature","submit_replication":"https://pith.science/pith/WTGFSDP2TC7HF3ESY7CSGJREF3/action/replication_record"}},"created_at":"2026-07-05T10:41:20.097121+00:00","updated_at":"2026-07-05T10:41:20.097121+00:00"}