{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:P7L6C64U334UV63NASGA2R462N","short_pith_number":"pith:P7L6C64U","schema_version":"1.0","canonical_sha256":"7fd7e17b94def94afb6d048c0d479ed342c3bc3b5ab72417ec9c41663ce6503f","source":{"kind":"arxiv","id":"2501.06645","version":3},"attestation_state":"computed","paper":{"title":"FocalPO: Enhancing Preference Optimizing by Focusing on Correct Preference Rankings","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Jindong Gu, Tong Liu, Volker Tresp, Wenxuan Zhou, Xiao Yu","submitted_at":"2025-01-11T21:41:27Z","abstract_excerpt":"Efficient preference optimization algorithms such as Direct Preference Optimization (DPO) have become a popular approach in aligning large language models (LLMs) with human preferences. These algorithms implicitly treat the LLM as a reward model, and focus on training it to correct misranked preference pairs. However, recent work~\\citep{chen2024preference} empirically finds that DPO training \\textit{rarely improves these misranked preference pairs}, despite its gradient emphasizing on these cases. We introduce FocalPO, a DPO variant that instead \\textit{down-weighs} misranked preference pairs "},"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":"2501.06645","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-11T21:41:27Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"1cc4dbcbcde760120290a5502dc8495841959fa756df9c3163fcba42e5f07c76","abstract_canon_sha256":"37e0f6efd8b321c7e4b1b79181dbbfb52525bc8f8270a88947b56aa443f8da51"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:44:00.418769Z","signature_b64":"ywvwazY+YqvevnEUPB9foDFpBPSw243GtNs1JeUVR8rrGx9Gbwevo6E9mrBimDq1pXa7k2PkXMnJORfn362rAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7fd7e17b94def94afb6d048c0d479ed342c3bc3b5ab72417ec9c41663ce6503f","last_reissued_at":"2026-07-05T11:44:00.418282Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:44:00.418282Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"FocalPO: Enhancing Preference Optimizing by Focusing on Correct Preference Rankings","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Jindong Gu, Tong Liu, Volker Tresp, Wenxuan Zhou, Xiao Yu","submitted_at":"2025-01-11T21:41:27Z","abstract_excerpt":"Efficient preference optimization algorithms such as Direct Preference Optimization (DPO) have become a popular approach in aligning large language models (LLMs) with human preferences. These algorithms implicitly treat the LLM as a reward model, and focus on training it to correct misranked preference pairs. However, recent work~\\citep{chen2024preference} empirically finds that DPO training \\textit{rarely improves these misranked preference pairs}, despite its gradient emphasizing on these cases. We introduce FocalPO, a DPO variant that instead \\textit{down-weighs} misranked preference pairs "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.06645","kind":"arxiv","version":3},"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/2501.06645/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":"2501.06645","created_at":"2026-07-05T11:44:00.418340+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.06645v3","created_at":"2026-07-05T11:44:00.418340+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.06645","created_at":"2026-07-05T11:44:00.418340+00:00"},{"alias_kind":"pith_short_12","alias_value":"P7L6C64U334U","created_at":"2026-07-05T11:44:00.418340+00:00"},{"alias_kind":"pith_short_16","alias_value":"P7L6C64U334UV63N","created_at":"2026-07-05T11:44:00.418340+00:00"},{"alias_kind":"pith_short_8","alias_value":"P7L6C64U","created_at":"2026-07-05T11:44:00.418340+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25906","citing_title":"OracleAnalyser: Analysing Implicit Semantics of Oracle Bone Scripts through MLLMs with Post-training","ref_index":31,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/P7L6C64U334UV63NASGA2R462N","json":"https://pith.science/pith/P7L6C64U334UV63NASGA2R462N.json","graph_json":"https://pith.science/api/pith-number/P7L6C64U334UV63NASGA2R462N/graph.json","events_json":"https://pith.science/api/pith-number/P7L6C64U334UV63NASGA2R462N/events.json","paper":"https://pith.science/paper/P7L6C64U"},"agent_actions":{"view_html":"https://pith.science/pith/P7L6C64U334UV63NASGA2R462N","download_json":"https://pith.science/pith/P7L6C64U334UV63NASGA2R462N.json","view_paper":"https://pith.science/paper/P7L6C64U","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.06645&json=true","fetch_graph":"https://pith.science/api/pith-number/P7L6C64U334UV63NASGA2R462N/graph.json","fetch_events":"https://pith.science/api/pith-number/P7L6C64U334UV63NASGA2R462N/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/P7L6C64U334UV63NASGA2R462N/action/timestamp_anchor","attest_storage":"https://pith.science/pith/P7L6C64U334UV63NASGA2R462N/action/storage_attestation","attest_author":"https://pith.science/pith/P7L6C64U334UV63NASGA2R462N/action/author_attestation","sign_citation":"https://pith.science/pith/P7L6C64U334UV63NASGA2R462N/action/citation_signature","submit_replication":"https://pith.science/pith/P7L6C64U334UV63NASGA2R462N/action/replication_record"}},"created_at":"2026-07-05T11:44:00.418340+00:00","updated_at":"2026-07-05T11:44:00.418340+00:00"}