{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:2JI7OU7GKMHUVEERPP5I5JZ42Q","short_pith_number":"pith:2JI7OU7G","schema_version":"1.0","canonical_sha256":"d251f753e6530f4a90917bfa8ea73cd41759b9c5d26e700f5f9f90dd4f0bdc09","source":{"kind":"arxiv","id":"2405.14953","version":5},"attestation_state":"computed","paper":{"title":"MallowsPO: Fine-Tune Your LLM with Preference Dispersions","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"David Yao, Hanyang Zhao, Haoxian Chen, Henry Lam, Wenpin Tang","submitted_at":"2024-05-23T18:01:11Z","abstract_excerpt":"Direct Preference Optimization (DPO) has recently emerged as a popular approach to improve reinforcement learning with human feedback (RLHF), leading to better techniques to fine-tune large language models (LLM). A weakness of DPO, however, lies in its lack of capability to characterize the diversity of human preferences. Inspired by Mallows' theory of preference ranking, we develop in this paper a new approach, the MallowsPO. A distinct feature of this approach is a dispersion index, which reflects the dispersion of human preference to prompts. We show that existing DPO models can be reduced "},"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":"2405.14953","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-05-23T18:01:11Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"9c377af5887133d4f53584a1bf438fa2fdd28f198a7bb8c6b8bce00ceb0f324c","abstract_canon_sha256":"8ef7eed017134e734d1977df226d7a57d53f1fdd628c324107e5b6b6fb3282af"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:50:42.209376Z","signature_b64":"Vc7bLxAleiPjBTkO/ETeNUdygXTf8xFSjqy0iDcCdnrAFOu2whesm2T0HCFE7V9SQtm+2Jv0awglODEV2eGxDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d251f753e6530f4a90917bfa8ea73cd41759b9c5d26e700f5f9f90dd4f0bdc09","last_reissued_at":"2026-07-05T10:50:42.208842Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:50:42.208842Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MallowsPO: Fine-Tune Your LLM with Preference Dispersions","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"David Yao, Hanyang Zhao, Haoxian Chen, Henry Lam, Wenpin Tang","submitted_at":"2024-05-23T18:01:11Z","abstract_excerpt":"Direct Preference Optimization (DPO) has recently emerged as a popular approach to improve reinforcement learning with human feedback (RLHF), leading to better techniques to fine-tune large language models (LLM). A weakness of DPO, however, lies in its lack of capability to characterize the diversity of human preferences. Inspired by Mallows' theory of preference ranking, we develop in this paper a new approach, the MallowsPO. A distinct feature of this approach is a dispersion index, which reflects the dispersion of human preference to prompts. We show that existing DPO models can be reduced "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.14953","kind":"arxiv","version":5},"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/2405.14953/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":"2405.14953","created_at":"2026-07-05T10:50:42.208904+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.14953v5","created_at":"2026-07-05T10:50:42.208904+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.14953","created_at":"2026-07-05T10:50:42.208904+00:00"},{"alias_kind":"pith_short_12","alias_value":"2JI7OU7GKMHU","created_at":"2026-07-05T10:50:42.208904+00:00"},{"alias_kind":"pith_short_16","alias_value":"2JI7OU7GKMHUVEER","created_at":"2026-07-05T10:50:42.208904+00:00"},{"alias_kind":"pith_short_8","alias_value":"2JI7OU7G","created_at":"2026-07-05T10:50:42.208904+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.18606","citing_title":"Steerable Cultural Preference Optimization of Reward Models","ref_index":22,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2JI7OU7GKMHUVEERPP5I5JZ42Q","json":"https://pith.science/pith/2JI7OU7GKMHUVEERPP5I5JZ42Q.json","graph_json":"https://pith.science/api/pith-number/2JI7OU7GKMHUVEERPP5I5JZ42Q/graph.json","events_json":"https://pith.science/api/pith-number/2JI7OU7GKMHUVEERPP5I5JZ42Q/events.json","paper":"https://pith.science/paper/2JI7OU7G"},"agent_actions":{"view_html":"https://pith.science/pith/2JI7OU7GKMHUVEERPP5I5JZ42Q","download_json":"https://pith.science/pith/2JI7OU7GKMHUVEERPP5I5JZ42Q.json","view_paper":"https://pith.science/paper/2JI7OU7G","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.14953&json=true","fetch_graph":"https://pith.science/api/pith-number/2JI7OU7GKMHUVEERPP5I5JZ42Q/graph.json","fetch_events":"https://pith.science/api/pith-number/2JI7OU7GKMHUVEERPP5I5JZ42Q/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2JI7OU7GKMHUVEERPP5I5JZ42Q/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2JI7OU7GKMHUVEERPP5I5JZ42Q/action/storage_attestation","attest_author":"https://pith.science/pith/2JI7OU7GKMHUVEERPP5I5JZ42Q/action/author_attestation","sign_citation":"https://pith.science/pith/2JI7OU7GKMHUVEERPP5I5JZ42Q/action/citation_signature","submit_replication":"https://pith.science/pith/2JI7OU7GKMHUVEERPP5I5JZ42Q/action/replication_record"}},"created_at":"2026-07-05T10:50:42.208904+00:00","updated_at":"2026-07-05T10:50:42.208904+00:00"}