{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:F6YF3DP2JL5F54GNILDRG7QY47","short_pith_number":"pith:F6YF3DP2","schema_version":"1.0","canonical_sha256":"2fb05d8dfa4afa5ef0cd42c7137e18e7fa9ea61d6ea556016029279adc46c2e7","source":{"kind":"arxiv","id":"2410.15287","version":1},"attestation_state":"computed","paper":{"title":"Training Language Models to Critique With Multi-agent Feedback","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chengqi Lyu, Chen Xu, Dahua Lin, Heyan Huang, Kai Chen, Shuaibin Li, Tian Lan, Wenwei Zhang, Xian-Ling Mao","submitted_at":"2024-10-20T04:57:45Z","abstract_excerpt":"Critique ability, a meta-cognitive capability of humans, presents significant challenges for LLMs to improve. Recent works primarily rely on supervised fine-tuning (SFT) using critiques generated by a single LLM like GPT-4. However, these model-generated critiques often exhibit flaws due to the inherent complexity of the critique. Consequently, fine-tuning LLMs on such flawed critiques typically limits the model's performance and propagates these flaws into the learned model. To overcome these challenges, this paper proposes a novel data generation pipeline, named MultiCritique, that improves "},"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":"2410.15287","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-10-20T04:57:45Z","cross_cats_sorted":[],"title_canon_sha256":"2701b794f3eb05136b74302095dc487157954d6e43a6537c99e5e62b6409b855","abstract_canon_sha256":"c873e32afa8ddb51a9a906fcb5009cda1ae61d13cb01eb7f7fe52adf5336f1e3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:23:17.191653Z","signature_b64":"M/SAtHpxd3RA8cazwLTGJZBAjEnUf1ZutfpYVmzw/GXS5a4g5A5J35lAXRtZsRtY3a6qATQ0C+QRNwoOOXc8Dg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2fb05d8dfa4afa5ef0cd42c7137e18e7fa9ea61d6ea556016029279adc46c2e7","last_reissued_at":"2026-07-05T09:23:17.191243Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:23:17.191243Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Training Language Models to Critique With Multi-agent Feedback","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chengqi Lyu, Chen Xu, Dahua Lin, Heyan Huang, Kai Chen, Shuaibin Li, Tian Lan, Wenwei Zhang, Xian-Ling Mao","submitted_at":"2024-10-20T04:57:45Z","abstract_excerpt":"Critique ability, a meta-cognitive capability of humans, presents significant challenges for LLMs to improve. Recent works primarily rely on supervised fine-tuning (SFT) using critiques generated by a single LLM like GPT-4. However, these model-generated critiques often exhibit flaws due to the inherent complexity of the critique. Consequently, fine-tuning LLMs on such flawed critiques typically limits the model's performance and propagates these flaws into the learned model. To overcome these challenges, this paper proposes a novel data generation pipeline, named MultiCritique, that improves "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.15287","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/2410.15287/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":"2410.15287","created_at":"2026-07-05T09:23:17.191297+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.15287v1","created_at":"2026-07-05T09:23:17.191297+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.15287","created_at":"2026-07-05T09:23:17.191297+00:00"},{"alias_kind":"pith_short_12","alias_value":"F6YF3DP2JL5F","created_at":"2026-07-05T09:23:17.191297+00:00"},{"alias_kind":"pith_short_16","alias_value":"F6YF3DP2JL5F54GN","created_at":"2026-07-05T09:23:17.191297+00:00"},{"alias_kind":"pith_short_8","alias_value":"F6YF3DP2","created_at":"2026-07-05T09:23:17.191297+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2411.08147","citing_title":"Large Language Models Can Self-Improve in Long-context Reasoning","ref_index":52,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/F6YF3DP2JL5F54GNILDRG7QY47","json":"https://pith.science/pith/F6YF3DP2JL5F54GNILDRG7QY47.json","graph_json":"https://pith.science/api/pith-number/F6YF3DP2JL5F54GNILDRG7QY47/graph.json","events_json":"https://pith.science/api/pith-number/F6YF3DP2JL5F54GNILDRG7QY47/events.json","paper":"https://pith.science/paper/F6YF3DP2"},"agent_actions":{"view_html":"https://pith.science/pith/F6YF3DP2JL5F54GNILDRG7QY47","download_json":"https://pith.science/pith/F6YF3DP2JL5F54GNILDRG7QY47.json","view_paper":"https://pith.science/paper/F6YF3DP2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.15287&json=true","fetch_graph":"https://pith.science/api/pith-number/F6YF3DP2JL5F54GNILDRG7QY47/graph.json","fetch_events":"https://pith.science/api/pith-number/F6YF3DP2JL5F54GNILDRG7QY47/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/F6YF3DP2JL5F54GNILDRG7QY47/action/timestamp_anchor","attest_storage":"https://pith.science/pith/F6YF3DP2JL5F54GNILDRG7QY47/action/storage_attestation","attest_author":"https://pith.science/pith/F6YF3DP2JL5F54GNILDRG7QY47/action/author_attestation","sign_citation":"https://pith.science/pith/F6YF3DP2JL5F54GNILDRG7QY47/action/citation_signature","submit_replication":"https://pith.science/pith/F6YF3DP2JL5F54GNILDRG7QY47/action/replication_record"}},"created_at":"2026-07-05T09:23:17.191297+00:00","updated_at":"2026-07-05T09:23:17.191297+00:00"}