{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:T3CPF6FMDSCLLXOAY3EJM76TLK","short_pith_number":"pith:T3CPF6FM","schema_version":"1.0","canonical_sha256":"9ec4f2f8ac1c84b5ddc0c6c8967fd35ab83ba8042c2c814224a228acb1b01862","source":{"kind":"arxiv","id":"2402.07314","version":3},"attestation_state":"computed","paper":{"title":"Online Iterative Reinforcement Learning from Human Feedback with General Preference Model","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Chenlu Ye, HanZe Dong, Nan Jiang, Tong Zhang, Wei Xiong, Yuheng Zhang","submitted_at":"2024-02-11T21:44:21Z","abstract_excerpt":"We investigate Reinforcement Learning from Human Feedback (RLHF) in the context of a general preference oracle. In particular, we do not assume the existence of a reward function and an oracle preference signal drawn from the Bradley-Terry model as most of the prior works do. We consider a standard mathematical formulation, the reverse-KL regularized minimax game between two LLMs for RLHF under general preference oracle. The learning objective of this formulation is to find a policy so that it is consistently preferred by the KL-regularized preference oracle over any competing LLMs. We show th"},"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":"2402.07314","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-02-11T21:44:21Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"6bc8a4e8b396505d047129dc2602cc5f31387418a2e15d2cf7ed85386fbd48e7","abstract_canon_sha256":"a0cf55c17006fac0f8939ca45261d7d5014dcccee3acd1f11273dfe336b6d5ff"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:34:11.423468Z","signature_b64":"cdWFKyGBgR5/ot7IqWRaG3bIdhuvga/sZYS/wr9SlOaipLd6Qu3SJlhMQd/mnabbkSd1PmJ6pivd+aXXD4YqAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9ec4f2f8ac1c84b5ddc0c6c8967fd35ab83ba8042c2c814224a228acb1b01862","last_reissued_at":"2026-07-05T09:34:11.422939Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:34:11.422939Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Online Iterative Reinforcement Learning from Human Feedback with General Preference Model","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Chenlu Ye, HanZe Dong, Nan Jiang, Tong Zhang, Wei Xiong, Yuheng Zhang","submitted_at":"2024-02-11T21:44:21Z","abstract_excerpt":"We investigate Reinforcement Learning from Human Feedback (RLHF) in the context of a general preference oracle. In particular, we do not assume the existence of a reward function and an oracle preference signal drawn from the Bradley-Terry model as most of the prior works do. We consider a standard mathematical formulation, the reverse-KL regularized minimax game between two LLMs for RLHF under general preference oracle. The learning objective of this formulation is to find a policy so that it is consistently preferred by the KL-regularized preference oracle over any competing LLMs. We show th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.07314","kind":"arxiv","version":3},"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/2402.07314/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":"2402.07314","created_at":"2026-07-05T09:34:11.423002+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.07314v3","created_at":"2026-07-05T09:34:11.423002+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.07314","created_at":"2026-07-05T09:34:11.423002+00:00"},{"alias_kind":"pith_short_12","alias_value":"T3CPF6FMDSCL","created_at":"2026-07-05T09:34:11.423002+00:00"},{"alias_kind":"pith_short_16","alias_value":"T3CPF6FMDSCLLXOA","created_at":"2026-07-05T09:34:11.423002+00:00"},{"alias_kind":"pith_short_8","alias_value":"T3CPF6FM","created_at":"2026-07-05T09:34:11.423002+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.06053","citing_title":"Online KL-Regularized Reinforcement Learning with Function Approximation under Misspecification","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2412.09413","citing_title":"Imitate, Explore, and Self-Improve: A Reproduction Report on Slow-thinking Reasoning Systems","ref_index":31,"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":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09214","citing_title":"Fast Rates for Offline Contextual Bandits with Forward-KL Regularization under Single-Policy Concentrability","ref_index":48,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/T3CPF6FMDSCLLXOAY3EJM76TLK","json":"https://pith.science/pith/T3CPF6FMDSCLLXOAY3EJM76TLK.json","graph_json":"https://pith.science/api/pith-number/T3CPF6FMDSCLLXOAY3EJM76TLK/graph.json","events_json":"https://pith.science/api/pith-number/T3CPF6FMDSCLLXOAY3EJM76TLK/events.json","paper":"https://pith.science/paper/T3CPF6FM"},"agent_actions":{"view_html":"https://pith.science/pith/T3CPF6FMDSCLLXOAY3EJM76TLK","download_json":"https://pith.science/pith/T3CPF6FMDSCLLXOAY3EJM76TLK.json","view_paper":"https://pith.science/paper/T3CPF6FM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.07314&json=true","fetch_graph":"https://pith.science/api/pith-number/T3CPF6FMDSCLLXOAY3EJM76TLK/graph.json","fetch_events":"https://pith.science/api/pith-number/T3CPF6FMDSCLLXOAY3EJM76TLK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/T3CPF6FMDSCLLXOAY3EJM76TLK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/T3CPF6FMDSCLLXOAY3EJM76TLK/action/storage_attestation","attest_author":"https://pith.science/pith/T3CPF6FMDSCLLXOAY3EJM76TLK/action/author_attestation","sign_citation":"https://pith.science/pith/T3CPF6FMDSCLLXOAY3EJM76TLK/action/citation_signature","submit_replication":"https://pith.science/pith/T3CPF6FMDSCLLXOAY3EJM76TLK/action/replication_record"}},"created_at":"2026-07-05T09:34:11.423002+00:00","updated_at":"2026-07-05T09:34:11.423002+00:00"}