{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ZEXJC47MQXT6A64IVZ7O4ONSKE","short_pith_number":"pith:ZEXJC47M","schema_version":"1.0","canonical_sha256":"c92e9173ec85e7e07b88ae7eee39b25130deb2a1cd874ba07c9ba151c14cadcb","source":{"kind":"arxiv","id":"2406.05882","version":1},"attestation_state":"computed","paper":{"title":"Distributional Preference Alignment of LLMs via Optimal Transport","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Apoorva Nitsure, Brian Belgodere, Igor Melnyk, Jerret Ross, Jiri Navratil, Kristjan Greenewald, Mattia Rigotti, Mikhail Yurochkin, Youssef Mroueh","submitted_at":"2024-06-09T18:41:05Z","abstract_excerpt":"Current LLM alignment techniques use pairwise human preferences at a sample level, and as such, they do not imply an alignment on the distributional level. We propose in this paper Alignment via Optimal Transport (AOT), a novel method for distributional preference alignment of LLMs. AOT aligns LLMs on unpaired preference data by making the reward distribution of the positive samples stochastically dominant in the first order on the distribution of negative samples. We introduce a convex relaxation of this first-order stochastic dominance and cast it as an optimal transport problem with a smoot"},"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":"2406.05882","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-09T18:41:05Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"7d9f2f63130a8a5d7ab470c0107152ee0916bd53548c85baa31910c6862191f9","abstract_canon_sha256":"960ec9b855dce0e4bb9b867559dd5c68adf6c1222bfe4cba3c4cea43fbed7dac"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:29:34.907750Z","signature_b64":"qKwJ2mQDxhH9BCVMPiMt01tHSRlkVLccg/vZowd4+vmJpc9iiGDsGlyUADUb11LFr5hF/jkZ4/EmBbQ4A9dLBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c92e9173ec85e7e07b88ae7eee39b25130deb2a1cd874ba07c9ba151c14cadcb","last_reissued_at":"2026-07-05T08:29:34.907287Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:29:34.907287Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Distributional Preference Alignment of LLMs via Optimal Transport","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Apoorva Nitsure, Brian Belgodere, Igor Melnyk, Jerret Ross, Jiri Navratil, Kristjan Greenewald, Mattia Rigotti, Mikhail Yurochkin, Youssef Mroueh","submitted_at":"2024-06-09T18:41:05Z","abstract_excerpt":"Current LLM alignment techniques use pairwise human preferences at a sample level, and as such, they do not imply an alignment on the distributional level. We propose in this paper Alignment via Optimal Transport (AOT), a novel method for distributional preference alignment of LLMs. AOT aligns LLMs on unpaired preference data by making the reward distribution of the positive samples stochastically dominant in the first order on the distribution of negative samples. We introduce a convex relaxation of this first-order stochastic dominance and cast it as an optimal transport problem with a smoot"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.05882","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/2406.05882/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":"2406.05882","created_at":"2026-07-05T08:29:34.907350+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.05882v1","created_at":"2026-07-05T08:29:34.907350+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.05882","created_at":"2026-07-05T08:29:34.907350+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZEXJC47MQXT6","created_at":"2026-07-05T08:29:34.907350+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZEXJC47MQXT6A64I","created_at":"2026-07-05T08:29:34.907350+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZEXJC47M","created_at":"2026-07-05T08:29:34.907350+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2502.16761","citing_title":"Language Model Fine-Tuning on Scaled Survey Data for Predicting Distributions of Public Opinions","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2411.10442","citing_title":"Enhancing the Reasoning Ability of Multimodal Large Language Models via Mixed Preference Optimization","ref_index":68,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08539","citing_title":"OpenVLThinkerV2: A Generalist Multimodal Reasoning Model for Multi-domain Visual Tasks","ref_index":25,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZEXJC47MQXT6A64IVZ7O4ONSKE","json":"https://pith.science/pith/ZEXJC47MQXT6A64IVZ7O4ONSKE.json","graph_json":"https://pith.science/api/pith-number/ZEXJC47MQXT6A64IVZ7O4ONSKE/graph.json","events_json":"https://pith.science/api/pith-number/ZEXJC47MQXT6A64IVZ7O4ONSKE/events.json","paper":"https://pith.science/paper/ZEXJC47M"},"agent_actions":{"view_html":"https://pith.science/pith/ZEXJC47MQXT6A64IVZ7O4ONSKE","download_json":"https://pith.science/pith/ZEXJC47MQXT6A64IVZ7O4ONSKE.json","view_paper":"https://pith.science/paper/ZEXJC47M","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.05882&json=true","fetch_graph":"https://pith.science/api/pith-number/ZEXJC47MQXT6A64IVZ7O4ONSKE/graph.json","fetch_events":"https://pith.science/api/pith-number/ZEXJC47MQXT6A64IVZ7O4ONSKE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZEXJC47MQXT6A64IVZ7O4ONSKE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZEXJC47MQXT6A64IVZ7O4ONSKE/action/storage_attestation","attest_author":"https://pith.science/pith/ZEXJC47MQXT6A64IVZ7O4ONSKE/action/author_attestation","sign_citation":"https://pith.science/pith/ZEXJC47MQXT6A64IVZ7O4ONSKE/action/citation_signature","submit_replication":"https://pith.science/pith/ZEXJC47MQXT6A64IVZ7O4ONSKE/action/replication_record"}},"created_at":"2026-07-05T08:29:34.907350+00:00","updated_at":"2026-07-05T08:29:34.907350+00:00"}