{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:T4G7BAVUINADSSMBR2HAEUUO3N","short_pith_number":"pith:T4G7BAVU","schema_version":"1.0","canonical_sha256":"9f0df082b443403949818e8e02528edb54eecf8242f721b98694ebba48149768","source":{"kind":"arxiv","id":"2407.00617","version":4},"attestation_state":"computed","paper":{"title":"Iterative Nash Policy Optimization: Aligning LLMs with General Preferences via No-Regret Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.GT"],"primary_cat":"cs.LG","authors_text":"Baolin Peng, Dian Yu, Dong Yu, Haitao Mi, Linfeng Song, Mingyue Huo, Nan Jiang, Ye Tian, Yuheng Zhang","submitted_at":"2024-06-30T08:00:34Z","abstract_excerpt":"Reinforcement Learning with Human Feedback (RLHF) has achieved great success in aligning large language models (LLMs) with human preferences. Prevalent RLHF approaches are reward-based, following the Bradley-Terry (BT) model assumption, which may not fully capture the complexity of human preferences. In this paper, we explore RLHF under a general preference framework and approach it from a game-theoretic perspective. Specifically, we formulate the problem as a two-player game and propose a novel online algorithm, iterative Nash policy optimization (INPO). The key idea is to let the policy play"},"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":"2407.00617","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-30T08:00:34Z","cross_cats_sorted":["cs.AI","cs.CL","cs.GT"],"title_canon_sha256":"72c9265a47dfa270d9dc1eedc98b412b7bc83cf20bc02c188d72c2b8ce0a5c3c","abstract_canon_sha256":"e3c299be26f75495348964205313a87ccd81c94b5f4f8d50e90db273f88fd3e6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:22:47.761391Z","signature_b64":"8GeVcuSurQZXf5ysZUHEtvhSfUjgZVu6lnRS+YS7rUmNnYZQcrofx9mw9XVZt24fi+OF7U06a8hcOJKydTq6DA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9f0df082b443403949818e8e02528edb54eecf8242f721b98694ebba48149768","last_reissued_at":"2026-07-05T10:22:47.758553Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:22:47.758553Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Iterative Nash Policy Optimization: Aligning LLMs with General Preferences via No-Regret Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.GT"],"primary_cat":"cs.LG","authors_text":"Baolin Peng, Dian Yu, Dong Yu, Haitao Mi, Linfeng Song, Mingyue Huo, Nan Jiang, Ye Tian, Yuheng Zhang","submitted_at":"2024-06-30T08:00:34Z","abstract_excerpt":"Reinforcement Learning with Human Feedback (RLHF) has achieved great success in aligning large language models (LLMs) with human preferences. Prevalent RLHF approaches are reward-based, following the Bradley-Terry (BT) model assumption, which may not fully capture the complexity of human preferences. In this paper, we explore RLHF under a general preference framework and approach it from a game-theoretic perspective. Specifically, we formulate the problem as a two-player game and propose a novel online algorithm, iterative Nash policy optimization (INPO). The key idea is to let the policy play"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.00617","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/2407.00617/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":"2407.00617","created_at":"2026-07-05T10:22:47.760477+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.00617v4","created_at":"2026-07-05T10:22:47.760477+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.00617","created_at":"2026-07-05T10:22:47.760477+00:00"},{"alias_kind":"pith_short_12","alias_value":"T4G7BAVUINAD","created_at":"2026-07-05T10:22:47.760477+00:00"},{"alias_kind":"pith_short_16","alias_value":"T4G7BAVUINADSSMB","created_at":"2026-07-05T10:22:47.760477+00:00"},{"alias_kind":"pith_short_8","alias_value":"T4G7BAVU","created_at":"2026-07-05T10:22:47.760477+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.13875","citing_title":"Common-agency Games for Multi-Objective Test-Time Alignment","ref_index":82,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09946","citing_title":"Structure from Strategic Interaction & Uncertainty: Risk Sensitive Games for Robust Preference Learning","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09363","citing_title":"Near-Optimal Last-Iterate Convergence for Zero-Sum Games with Bandit Feedback and Opponent Actions","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09946","citing_title":"Structure from Strategic Interaction & Uncertainty: Risk Sensitive Games for Robust Preference Learning","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04494","citing_title":"Towards General Preference Alignment: Diffusion Models at Nash Equilibrium","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07331","citing_title":"Rethinking Importance Sampling in LLM Policy Optimization: A Cumulative Token Perspective","ref_index":26,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/T4G7BAVUINADSSMBR2HAEUUO3N","json":"https://pith.science/pith/T4G7BAVUINADSSMBR2HAEUUO3N.json","graph_json":"https://pith.science/api/pith-number/T4G7BAVUINADSSMBR2HAEUUO3N/graph.json","events_json":"https://pith.science/api/pith-number/T4G7BAVUINADSSMBR2HAEUUO3N/events.json","paper":"https://pith.science/paper/T4G7BAVU"},"agent_actions":{"view_html":"https://pith.science/pith/T4G7BAVUINADSSMBR2HAEUUO3N","download_json":"https://pith.science/pith/T4G7BAVUINADSSMBR2HAEUUO3N.json","view_paper":"https://pith.science/paper/T4G7BAVU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.00617&json=true","fetch_graph":"https://pith.science/api/pith-number/T4G7BAVUINADSSMBR2HAEUUO3N/graph.json","fetch_events":"https://pith.science/api/pith-number/T4G7BAVUINADSSMBR2HAEUUO3N/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/T4G7BAVUINADSSMBR2HAEUUO3N/action/timestamp_anchor","attest_storage":"https://pith.science/pith/T4G7BAVUINADSSMBR2HAEUUO3N/action/storage_attestation","attest_author":"https://pith.science/pith/T4G7BAVUINADSSMBR2HAEUUO3N/action/author_attestation","sign_citation":"https://pith.science/pith/T4G7BAVUINADSSMBR2HAEUUO3N/action/citation_signature","submit_replication":"https://pith.science/pith/T4G7BAVUINADSSMBR2HAEUUO3N/action/replication_record"}},"created_at":"2026-07-05T10:22:47.760477+00:00","updated_at":"2026-07-05T10:22:47.760477+00:00"}