{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:NOGWMIK3NMVKFBFJRTEG5VLGAS","short_pith_number":"pith:NOGWMIK3","schema_version":"1.0","canonical_sha256":"6b8d66215b6b2aa284a98cc86ed56604ab6ab9e768cb1633d116e9fd8c94ee4e","source":{"kind":"arxiv","id":"2505.23749","version":1},"attestation_state":"computed","paper":{"title":"Distortion of AI Alignment: Does Preference Optimization Optimize for Preferences?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.GT"],"primary_cat":"cs.LG","authors_text":"Kunhe Yang, Nika Haghtalab, Paul G\\\"olz","submitted_at":"2025-05-29T17:59:20Z","abstract_excerpt":"After pre-training, large language models are aligned with human preferences based on pairwise comparisons. State-of-the-art alignment methods (such as PPO-based RLHF and DPO) are built on the assumption of aligning with a single preference model, despite being deployed in settings where users have diverse preferences. As a result, it is not even clear that these alignment methods produce models that satisfy users on average -- a minimal requirement for pluralistic alignment. Drawing on social choice theory and modeling users' comparisons through individual Bradley-Terry (BT) models, we introd"},"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":"2505.23749","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-05-29T17:59:20Z","cross_cats_sorted":["cs.GT"],"title_canon_sha256":"07cdcfaecfad9ea176b12ed57117319d0e9fab6ce2c3935753c6e84f251f3fca","abstract_canon_sha256":"d3fc61e9afd3a6b7499b2c0b1c99f1140d0982f5c526eb4715e08dacfcf52b0c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:12:13.624749Z","signature_b64":"4AbGDiZCSTij+enpOY+A+OthRjxHNzhrqsYScMaHLK8/5qIuSIrVh1M97y8fRr+aS/Rp35dfENThgcgL2S9PAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6b8d66215b6b2aa284a98cc86ed56604ab6ab9e768cb1633d116e9fd8c94ee4e","last_reissued_at":"2026-07-05T11:12:13.624267Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:12:13.624267Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Distortion of AI Alignment: Does Preference Optimization Optimize for Preferences?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.GT"],"primary_cat":"cs.LG","authors_text":"Kunhe Yang, Nika Haghtalab, Paul G\\\"olz","submitted_at":"2025-05-29T17:59:20Z","abstract_excerpt":"After pre-training, large language models are aligned with human preferences based on pairwise comparisons. State-of-the-art alignment methods (such as PPO-based RLHF and DPO) are built on the assumption of aligning with a single preference model, despite being deployed in settings where users have diverse preferences. As a result, it is not even clear that these alignment methods produce models that satisfy users on average -- a minimal requirement for pluralistic alignment. Drawing on social choice theory and modeling users' comparisons through individual Bradley-Terry (BT) models, we introd"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.23749","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/2505.23749/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":"2505.23749","created_at":"2026-07-05T11:12:13.624322+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.23749v1","created_at":"2026-07-05T11:12:13.624322+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.23749","created_at":"2026-07-05T11:12:13.624322+00:00"},{"alias_kind":"pith_short_12","alias_value":"NOGWMIK3NMVK","created_at":"2026-07-05T11:12:13.624322+00:00"},{"alias_kind":"pith_short_16","alias_value":"NOGWMIK3NMVKFBFJ","created_at":"2026-07-05T11:12:13.624322+00:00"},{"alias_kind":"pith_short_8","alias_value":"NOGWMIK3","created_at":"2026-07-05T11:12:13.624322+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.27733","citing_title":"Mind the Gap: Structure-Aware Consistency in Preference Learning","ref_index":23,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NOGWMIK3NMVKFBFJRTEG5VLGAS","json":"https://pith.science/pith/NOGWMIK3NMVKFBFJRTEG5VLGAS.json","graph_json":"https://pith.science/api/pith-number/NOGWMIK3NMVKFBFJRTEG5VLGAS/graph.json","events_json":"https://pith.science/api/pith-number/NOGWMIK3NMVKFBFJRTEG5VLGAS/events.json","paper":"https://pith.science/paper/NOGWMIK3"},"agent_actions":{"view_html":"https://pith.science/pith/NOGWMIK3NMVKFBFJRTEG5VLGAS","download_json":"https://pith.science/pith/NOGWMIK3NMVKFBFJRTEG5VLGAS.json","view_paper":"https://pith.science/paper/NOGWMIK3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.23749&json=true","fetch_graph":"https://pith.science/api/pith-number/NOGWMIK3NMVKFBFJRTEG5VLGAS/graph.json","fetch_events":"https://pith.science/api/pith-number/NOGWMIK3NMVKFBFJRTEG5VLGAS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NOGWMIK3NMVKFBFJRTEG5VLGAS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NOGWMIK3NMVKFBFJRTEG5VLGAS/action/storage_attestation","attest_author":"https://pith.science/pith/NOGWMIK3NMVKFBFJRTEG5VLGAS/action/author_attestation","sign_citation":"https://pith.science/pith/NOGWMIK3NMVKFBFJRTEG5VLGAS/action/citation_signature","submit_replication":"https://pith.science/pith/NOGWMIK3NMVKFBFJRTEG5VLGAS/action/replication_record"}},"created_at":"2026-07-05T11:12:13.624322+00:00","updated_at":"2026-07-05T11:12:13.624322+00:00"}