{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:FZGJFDIMHFDON4T4LGI5BFI523","short_pith_number":"pith:FZGJFDIM","canonical_record":{"source":{"id":"2406.15568","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-21T18:06:30Z","cross_cats_sorted":[],"title_canon_sha256":"4dc639ba0cde351e1ccde57a18ec3d9e0cf4d66187a456d3b6aef028ddff044d","abstract_canon_sha256":"9029ef5b42a5f4e875cad6c8dc2f9944099102f216e0d7cab3ccfcfe762088be"},"schema_version":"1.0"},"canonical_sha256":"2e4c928d0c3946e6f27c5991d0951dd6dee40f8f906c9c31d7c3b986e678d589","source":{"kind":"arxiv","id":"2406.15568","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.15568","created_at":"2026-07-05T08:41:45Z"},{"alias_kind":"arxiv_version","alias_value":"2406.15568v2","created_at":"2026-07-05T08:41:45Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.15568","created_at":"2026-07-05T08:41:45Z"},{"alias_kind":"pith_short_12","alias_value":"FZGJFDIMHFDO","created_at":"2026-07-05T08:41:45Z"},{"alias_kind":"pith_short_16","alias_value":"FZGJFDIMHFDON4T4","created_at":"2026-07-05T08:41:45Z"},{"alias_kind":"pith_short_8","alias_value":"FZGJFDIM","created_at":"2026-07-05T08:41:45Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:FZGJFDIMHFDON4T4LGI5BFI523","target":"record","payload":{"canonical_record":{"source":{"id":"2406.15568","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-21T18:06:30Z","cross_cats_sorted":[],"title_canon_sha256":"4dc639ba0cde351e1ccde57a18ec3d9e0cf4d66187a456d3b6aef028ddff044d","abstract_canon_sha256":"9029ef5b42a5f4e875cad6c8dc2f9944099102f216e0d7cab3ccfcfe762088be"},"schema_version":"1.0"},"canonical_sha256":"2e4c928d0c3946e6f27c5991d0951dd6dee40f8f906c9c31d7c3b986e678d589","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:41:45.856786Z","signature_b64":"3y9eAEVrZrrPPs8wYN4IDZnmenf4XV/NwMRxTnf5RTVNv2yv9dLddtnRx1Xsbd5vPLda898Wg5ugzQvp0pouCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2e4c928d0c3946e6f27c5991d0951dd6dee40f8f906c9c31d7c3b986e678d589","last_reissued_at":"2026-07-05T08:41:45.856370Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:41:45.856370Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2406.15568","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T08:41:45Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"cAsIMv2MNyZ5R09NTY4T16bQZAJbIa7PuhLlmCGQu22jATkud+SVSwSVg0rbAznP0W2TFFeemztL/y8mAfv6AA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-08T17:57:11.255842Z"},"content_sha256":"eae002cc64c646cbfb9b8aaa4d188a248d46ec2bf2bf93747c2638b68d5c5db9","schema_version":"1.0","event_id":"sha256:eae002cc64c646cbfb9b8aaa4d188a248d46ec2bf2bf93747c2638b68d5c5db9"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:FZGJFDIMHFDON4T4LGI5BFI523","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Robust Reinforcement Learning from Corrupted Human Feedback","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Alexander Bukharin, Haoming Jiang, Ilgee Hong, Qingru Zhang, Tuo Zhao, Zichong Li, Zixuan Zhang","submitted_at":"2024-06-21T18:06:30Z","abstract_excerpt":"Reinforcement learning from human feedback (RLHF) provides a principled framework for aligning AI systems with human preference data. For various reasons, e.g., personal bias, context ambiguity, lack of training, etc, human annotators may give incorrect or inconsistent preference labels. To tackle this challenge, we propose a robust RLHF approach -- $R^3M$, which models the potentially corrupted preference label as sparse outliers. Accordingly, we formulate the robust reward learning as an $\\ell_1$-regularized maximum likelihood estimation problem. Computationally, we develop an efficient alte"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.15568","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/2406.15568/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T08:41:45Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"3VK2LjXqMIiq6NJpPLApGrij064rbAKyAJSFBcHb/DFlRcl7jORVUox5Sc7swtpDchDvhHmg4G+npWowCk0yCA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-08T17:57:11.256435Z"},"content_sha256":"8d2dc7aba9e3ebb636646bfcb942c0b14e0f87bc0929c3bd8fa7f2f9366606ae","schema_version":"1.0","event_id":"sha256:8d2dc7aba9e3ebb636646bfcb942c0b14e0f87bc0929c3bd8fa7f2f9366606ae"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/FZGJFDIMHFDON4T4LGI5BFI523/bundle.json","state_url":"https://pith.science/pith/FZGJFDIMHFDON4T4LGI5BFI523/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/FZGJFDIMHFDON4T4LGI5BFI523/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-08T17:57:11Z","links":{"resolver":"https://pith.science/pith/FZGJFDIMHFDON4T4LGI5BFI523","bundle":"https://pith.science/pith/FZGJFDIMHFDON4T4LGI5BFI523/bundle.json","state":"https://pith.science/pith/FZGJFDIMHFDON4T4LGI5BFI523/state.json","well_known_bundle":"https://pith.science/.well-known/pith/FZGJFDIMHFDON4T4LGI5BFI523/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:FZGJFDIMHFDON4T4LGI5BFI523","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"9029ef5b42a5f4e875cad6c8dc2f9944099102f216e0d7cab3ccfcfe762088be","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-21T18:06:30Z","title_canon_sha256":"4dc639ba0cde351e1ccde57a18ec3d9e0cf4d66187a456d3b6aef028ddff044d"},"schema_version":"1.0","source":{"id":"2406.15568","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.15568","created_at":"2026-07-05T08:41:45Z"},{"alias_kind":"arxiv_version","alias_value":"2406.15568v2","created_at":"2026-07-05T08:41:45Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.15568","created_at":"2026-07-05T08:41:45Z"},{"alias_kind":"pith_short_12","alias_value":"FZGJFDIMHFDO","created_at":"2026-07-05T08:41:45Z"},{"alias_kind":"pith_short_16","alias_value":"FZGJFDIMHFDON4T4","created_at":"2026-07-05T08:41:45Z"},{"alias_kind":"pith_short_8","alias_value":"FZGJFDIM","created_at":"2026-07-05T08:41:45Z"}],"graph_snapshots":[{"event_id":"sha256:8d2dc7aba9e3ebb636646bfcb942c0b14e0f87bc0929c3bd8fa7f2f9366606ae","target":"graph","created_at":"2026-07-05T08:41:45Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2406.15568/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reinforcement learning from human feedback (RLHF) provides a principled framework for aligning AI systems with human preference data. For various reasons, e.g., personal bias, context ambiguity, lack of training, etc, human annotators may give incorrect or inconsistent preference labels. To tackle this challenge, we propose a robust RLHF approach -- $R^3M$, which models the potentially corrupted preference label as sparse outliers. Accordingly, we formulate the robust reward learning as an $\\ell_1$-regularized maximum likelihood estimation problem. Computationally, we develop an efficient alte","authors_text":"Alexander Bukharin, Haoming Jiang, Ilgee Hong, Qingru Zhang, Tuo Zhao, Zichong Li, Zixuan Zhang","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-21T18:06:30Z","title":"Robust Reinforcement Learning from Corrupted Human Feedback"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.15568","kind":"arxiv","version":2},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:eae002cc64c646cbfb9b8aaa4d188a248d46ec2bf2bf93747c2638b68d5c5db9","target":"record","created_at":"2026-07-05T08:41:45Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"9029ef5b42a5f4e875cad6c8dc2f9944099102f216e0d7cab3ccfcfe762088be","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-21T18:06:30Z","title_canon_sha256":"4dc639ba0cde351e1ccde57a18ec3d9e0cf4d66187a456d3b6aef028ddff044d"},"schema_version":"1.0","source":{"id":"2406.15568","kind":"arxiv","version":2}},"canonical_sha256":"2e4c928d0c3946e6f27c5991d0951dd6dee40f8f906c9c31d7c3b986e678d589","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"2e4c928d0c3946e6f27c5991d0951dd6dee40f8f906c9c31d7c3b986e678d589","first_computed_at":"2026-07-05T08:41:45.856370Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:41:45.856370Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"3y9eAEVrZrrPPs8wYN4IDZnmenf4XV/NwMRxTnf5RTVNv2yv9dLddtnRx1Xsbd5vPLda898Wg5ugzQvp0pouCA==","signature_status":"signed_v1","signed_at":"2026-07-05T08:41:45.856786Z","signed_message":"canonical_sha256_bytes"},"source_id":"2406.15568","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:eae002cc64c646cbfb9b8aaa4d188a248d46ec2bf2bf93747c2638b68d5c5db9","sha256:8d2dc7aba9e3ebb636646bfcb942c0b14e0f87bc0929c3bd8fa7f2f9366606ae"],"state_sha256":"af46710e5566579bb9f55b9e302dcdb60eb399b69c19a45d84264dea1b118ff0"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"zvczVBbY2b0iycWOX22jYV8XXz2vpWKuYwmHyVlcu8fQMoDCb24+9C7rHqMbZmwwYJFRhKEw8yNvmz5Ln1PKDw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-08T17:57:11.261241Z","bundle_sha256":"31f25dabc89e3d6bdfca3d61f7ba11a4b5ea6f72b1b3dacbfad588ce3d3d0224"}}