{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:BAPKPM2A7PECQXIUJ2GNDZ3ILG","short_pith_number":"pith:BAPKPM2A","schema_version":"1.0","canonical_sha256":"081ea7b340fbc8285d144e8cd1e76859a3df7156ccf9fd59c8abc886c6fc39db","source":{"kind":"arxiv","id":"2406.01013","version":2},"attestation_state":"computed","paper":{"title":"Scalable Ensembling For Mitigating Reward Overoptimisation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Ahmed M. Ahmed, Rafael Rafailov, Sanmi Koyejo, Stepan Sharkov, Xuechen Li","submitted_at":"2024-06-03T05:46:53Z","abstract_excerpt":"Reinforcement Learning from Human Feedback (RLHF) has enabled significant advancements within language modeling for powerful, instruction-following models. However, the alignment of these models remains a pressing challenge as the policy tends to overfit the learned ``proxy\" reward model past an inflection point of utility as measured by a ``gold\" reward model that is more performant -- a phenomenon known as overoptimisation. Prior work has mitigated this issue by computing a pessimistic statistic over an ensemble of reward models, which is common in Offline Reinforcement Learning but incredib"},"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":"2406.01013","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-03T05:46:53Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"d74ff69db6e93de39b571e17fa53610dedafa980d90cf5628bb600d0cf0fb5f8","abstract_canon_sha256":"671bd616d612963c99f15394ca187ce16429a5b3842c8accfe57b3c3aeee3565"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:34:19.966174Z","signature_b64":"k2vBDLnkxHij6SM1BDWHEsR1TuMEh3PhtzpXeN0WsJViDkrebYVkrXgSFcygP62QbQzKMpoEcACCANTfmKynBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"081ea7b340fbc8285d144e8cd1e76859a3df7156ccf9fd59c8abc886c6fc39db","last_reissued_at":"2026-07-05T08:34:19.965720Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:34:19.965720Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scalable Ensembling For Mitigating Reward Overoptimisation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Ahmed M. Ahmed, Rafael Rafailov, Sanmi Koyejo, Stepan Sharkov, Xuechen Li","submitted_at":"2024-06-03T05:46:53Z","abstract_excerpt":"Reinforcement Learning from Human Feedback (RLHF) has enabled significant advancements within language modeling for powerful, instruction-following models. However, the alignment of these models remains a pressing challenge as the policy tends to overfit the learned ``proxy\" reward model past an inflection point of utility as measured by a ``gold\" reward model that is more performant -- a phenomenon known as overoptimisation. Prior work has mitigated this issue by computing a pessimistic statistic over an ensemble of reward models, which is common in Offline Reinforcement Learning but incredib"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.01013","kind":"arxiv","version":2},"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/2406.01013/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":"2406.01013","created_at":"2026-07-05T08:34:19.965790+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.01013v2","created_at":"2026-07-05T08:34:19.965790+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.01013","created_at":"2026-07-05T08:34:19.965790+00:00"},{"alias_kind":"pith_short_12","alias_value":"BAPKPM2A7PEC","created_at":"2026-07-05T08:34:19.965790+00:00"},{"alias_kind":"pith_short_16","alias_value":"BAPKPM2A7PECQXIU","created_at":"2026-07-05T08:34:19.965790+00:00"},{"alias_kind":"pith_short_8","alias_value":"BAPKPM2A","created_at":"2026-07-05T08:34:19.965790+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.07105","citing_title":"Theoretical Limits of Language Model Alignment","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BAPKPM2A7PECQXIUJ2GNDZ3ILG","json":"https://pith.science/pith/BAPKPM2A7PECQXIUJ2GNDZ3ILG.json","graph_json":"https://pith.science/api/pith-number/BAPKPM2A7PECQXIUJ2GNDZ3ILG/graph.json","events_json":"https://pith.science/api/pith-number/BAPKPM2A7PECQXIUJ2GNDZ3ILG/events.json","paper":"https://pith.science/paper/BAPKPM2A"},"agent_actions":{"view_html":"https://pith.science/pith/BAPKPM2A7PECQXIUJ2GNDZ3ILG","download_json":"https://pith.science/pith/BAPKPM2A7PECQXIUJ2GNDZ3ILG.json","view_paper":"https://pith.science/paper/BAPKPM2A","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.01013&json=true","fetch_graph":"https://pith.science/api/pith-number/BAPKPM2A7PECQXIUJ2GNDZ3ILG/graph.json","fetch_events":"https://pith.science/api/pith-number/BAPKPM2A7PECQXIUJ2GNDZ3ILG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BAPKPM2A7PECQXIUJ2GNDZ3ILG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BAPKPM2A7PECQXIUJ2GNDZ3ILG/action/storage_attestation","attest_author":"https://pith.science/pith/BAPKPM2A7PECQXIUJ2GNDZ3ILG/action/author_attestation","sign_citation":"https://pith.science/pith/BAPKPM2A7PECQXIUJ2GNDZ3ILG/action/citation_signature","submit_replication":"https://pith.science/pith/BAPKPM2A7PECQXIUJ2GNDZ3ILG/action/replication_record"}},"created_at":"2026-07-05T08:34:19.965790+00:00","updated_at":"2026-07-05T08:34:19.965790+00:00"}