{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:U4KOCJSVJKC7BVZ2JJSJU7ZJTN","short_pith_number":"pith:U4KOCJSV","schema_version":"1.0","canonical_sha256":"a714e126554a85f0d73a4a649a7f299b574cd885cdb3adc2380e7140587025ba","source":{"kind":"arxiv","id":"2506.12350","version":1},"attestation_state":"computed","paper":{"title":"Theoretical Tensions in RLHF: Reconciling Empirical Success with Inconsistencies in Social Choice Theory","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"stat.ML","authors_text":"Jiancong Xiao, Kaizhao Liu, Qi Long, Weijie J. Su, Zhekun Shi","submitted_at":"2025-06-14T05:14:49Z","abstract_excerpt":"Despite its empirical success, Reinforcement Learning from Human Feedback (RLHF) has been shown to violate almost all the fundamental axioms in social choice theory -- such as majority consistency, pairwise majority consistency, and Condorcet consistency. This raises a foundational question: why does RLHF perform so well in practice if it fails these seemingly essential properties? In this paper, we resolve this paradox by showing that under mild and empirically plausible assumptions on the preference profile, RLHF does satisfy pairwise majority and Condorcet consistency. These assumptions are"},"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":"2506.12350","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2025-06-14T05:14:49Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"b91832a99a95a6a51e52fa40fd7a61cf384840e2a51687045b462f5592d96cd0","abstract_canon_sha256":"6328d9bf6b852c8396e9620334e61dcb18b6e4b78a2d0e5737e498bd05b83f1f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:21:51.756918Z","signature_b64":"brdEkLwaufiZIrSklQAVBynkJOoMXM500E2UVIC7LORc8ytJbldyEnxnF3sd3+8Yu7gNVvD1IoY5Qd49cqByBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a714e126554a85f0d73a4a649a7f299b574cd885cdb3adc2380e7140587025ba","last_reissued_at":"2026-07-05T11:21:51.756410Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:21:51.756410Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Theoretical Tensions in RLHF: Reconciling Empirical Success with Inconsistencies in Social Choice Theory","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"stat.ML","authors_text":"Jiancong Xiao, Kaizhao Liu, Qi Long, Weijie J. Su, Zhekun Shi","submitted_at":"2025-06-14T05:14:49Z","abstract_excerpt":"Despite its empirical success, Reinforcement Learning from Human Feedback (RLHF) has been shown to violate almost all the fundamental axioms in social choice theory -- such as majority consistency, pairwise majority consistency, and Condorcet consistency. This raises a foundational question: why does RLHF perform so well in practice if it fails these seemingly essential properties? In this paper, we resolve this paradox by showing that under mild and empirically plausible assumptions on the preference profile, RLHF does satisfy pairwise majority and Condorcet consistency. These assumptions are"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.12350","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/2506.12350/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":"2506.12350","created_at":"2026-07-05T11:21:51.756467+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.12350v1","created_at":"2026-07-05T11:21:51.756467+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.12350","created_at":"2026-07-05T11:21:51.756467+00:00"},{"alias_kind":"pith_short_12","alias_value":"U4KOCJSVJKC7","created_at":"2026-07-05T11:21:51.756467+00:00"},{"alias_kind":"pith_short_16","alias_value":"U4KOCJSVJKC7BVZ2","created_at":"2026-07-05T11:21:51.756467+00:00"},{"alias_kind":"pith_short_8","alias_value":"U4KOCJSV","created_at":"2026-07-05T11:21:51.756467+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21550","citing_title":"AI Alignment From Social Choice Perspectives","ref_index":81,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02340","citing_title":"Transitivity in Inhomogeneous Random Tournaments","ref_index":51,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/U4KOCJSVJKC7BVZ2JJSJU7ZJTN","json":"https://pith.science/pith/U4KOCJSVJKC7BVZ2JJSJU7ZJTN.json","graph_json":"https://pith.science/api/pith-number/U4KOCJSVJKC7BVZ2JJSJU7ZJTN/graph.json","events_json":"https://pith.science/api/pith-number/U4KOCJSVJKC7BVZ2JJSJU7ZJTN/events.json","paper":"https://pith.science/paper/U4KOCJSV"},"agent_actions":{"view_html":"https://pith.science/pith/U4KOCJSVJKC7BVZ2JJSJU7ZJTN","download_json":"https://pith.science/pith/U4KOCJSVJKC7BVZ2JJSJU7ZJTN.json","view_paper":"https://pith.science/paper/U4KOCJSV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.12350&json=true","fetch_graph":"https://pith.science/api/pith-number/U4KOCJSVJKC7BVZ2JJSJU7ZJTN/graph.json","fetch_events":"https://pith.science/api/pith-number/U4KOCJSVJKC7BVZ2JJSJU7ZJTN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/U4KOCJSVJKC7BVZ2JJSJU7ZJTN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/U4KOCJSVJKC7BVZ2JJSJU7ZJTN/action/storage_attestation","attest_author":"https://pith.science/pith/U4KOCJSVJKC7BVZ2JJSJU7ZJTN/action/author_attestation","sign_citation":"https://pith.science/pith/U4KOCJSVJKC7BVZ2JJSJU7ZJTN/action/citation_signature","submit_replication":"https://pith.science/pith/U4KOCJSVJKC7BVZ2JJSJU7ZJTN/action/replication_record"}},"created_at":"2026-07-05T11:21:51.756467+00:00","updated_at":"2026-07-05T11:21:51.756467+00:00"}