{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:5NNTALUTGGHLM3QFZLLAZPY6HZ","short_pith_number":"pith:5NNTALUT","schema_version":"1.0","canonical_sha256":"eb5b302e93318eb66e05cad60cbf1e3e5ad8debb96df7dbea80ed9fe64fbde5d","source":{"kind":"arxiv","id":"2405.17956","version":4},"attestation_state":"computed","paper":{"title":"Unified Preference Optimization: Language Model Alignment Beyond the Preference Frontier","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Anirudhan Badrinath, Jiajing Xu, Prabhat Agarwal","submitted_at":"2024-05-28T08:35:48Z","abstract_excerpt":"For aligning large language models (LLMs), prior work has leveraged reinforcement learning via human feedback (RLHF) or variations of direct preference optimization (DPO). While DPO offers a simpler framework based on maximum likelihood estimation, it compromises on the ability to easily tune language models to maximize auxiliary, non-preferential objectives according to the LLM designer's preferences (e.g., tuning lexical style or minimizing specific kinds of harmful content). Critically, these designer objectives may not be amply human-labeled or represented in available data, align with use"},"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":"2405.17956","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-05-28T08:35:48Z","cross_cats_sorted":[],"title_canon_sha256":"665ff312c9e08987a0e91d923c89d071bebdccc763c37000ac9708e608e0cde1","abstract_canon_sha256":"612dc8f92a54d99f74197f617859d18320ea49f5b8997dfa1a6c47dfa6f3bbda"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:08:57.711365Z","signature_b64":"FNB7INdwi/THQwdZDSFHmR98PGE5WDDOEGueYH16QN4vNqsKBLcsxYo1/Cdnrnx9aKmX57XfcSGD1q4mJro6AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"eb5b302e93318eb66e05cad60cbf1e3e5ad8debb96df7dbea80ed9fe64fbde5d","last_reissued_at":"2026-07-05T11:08:57.710885Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:08:57.710885Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Unified Preference Optimization: Language Model Alignment Beyond the Preference Frontier","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Anirudhan Badrinath, Jiajing Xu, Prabhat Agarwal","submitted_at":"2024-05-28T08:35:48Z","abstract_excerpt":"For aligning large language models (LLMs), prior work has leveraged reinforcement learning via human feedback (RLHF) or variations of direct preference optimization (DPO). While DPO offers a simpler framework based on maximum likelihood estimation, it compromises on the ability to easily tune language models to maximize auxiliary, non-preferential objectives according to the LLM designer's preferences (e.g., tuning lexical style or minimizing specific kinds of harmful content). Critically, these designer objectives may not be amply human-labeled or represented in available data, align with use"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.17956","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/2405.17956/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":"2405.17956","created_at":"2026-07-05T11:08:57.710943+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.17956v4","created_at":"2026-07-05T11:08:57.710943+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.17956","created_at":"2026-07-05T11:08:57.710943+00:00"},{"alias_kind":"pith_short_12","alias_value":"5NNTALUTGGHL","created_at":"2026-07-05T11:08:57.710943+00:00"},{"alias_kind":"pith_short_16","alias_value":"5NNTALUTGGHLM3QF","created_at":"2026-07-05T11:08:57.710943+00:00"},{"alias_kind":"pith_short_8","alias_value":"5NNTALUT","created_at":"2026-07-05T11:08:57.710943+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2506.07165","citing_title":"AMoPO: Adaptive Multi-objective Preference Optimization without Reward Models and Reference Models","ref_index":3,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5NNTALUTGGHLM3QFZLLAZPY6HZ","json":"https://pith.science/pith/5NNTALUTGGHLM3QFZLLAZPY6HZ.json","graph_json":"https://pith.science/api/pith-number/5NNTALUTGGHLM3QFZLLAZPY6HZ/graph.json","events_json":"https://pith.science/api/pith-number/5NNTALUTGGHLM3QFZLLAZPY6HZ/events.json","paper":"https://pith.science/paper/5NNTALUT"},"agent_actions":{"view_html":"https://pith.science/pith/5NNTALUTGGHLM3QFZLLAZPY6HZ","download_json":"https://pith.science/pith/5NNTALUTGGHLM3QFZLLAZPY6HZ.json","view_paper":"https://pith.science/paper/5NNTALUT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.17956&json=true","fetch_graph":"https://pith.science/api/pith-number/5NNTALUTGGHLM3QFZLLAZPY6HZ/graph.json","fetch_events":"https://pith.science/api/pith-number/5NNTALUTGGHLM3QFZLLAZPY6HZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5NNTALUTGGHLM3QFZLLAZPY6HZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5NNTALUTGGHLM3QFZLLAZPY6HZ/action/storage_attestation","attest_author":"https://pith.science/pith/5NNTALUTGGHLM3QFZLLAZPY6HZ/action/author_attestation","sign_citation":"https://pith.science/pith/5NNTALUTGGHLM3QFZLLAZPY6HZ/action/citation_signature","submit_replication":"https://pith.science/pith/5NNTALUTGGHLM3QFZLLAZPY6HZ/action/replication_record"}},"created_at":"2026-07-05T11:08:57.710943+00:00","updated_at":"2026-07-05T11:08:57.710943+00:00"}