{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:OQNNOHGKQUBKFOQHM3DWCU7YYU","short_pith_number":"pith:OQNNOHGK","canonical_record":{"source":{"id":"2301.11270","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-01-26T18:07:21Z","cross_cats_sorted":["cs.AI","cs.HC","math.ST","stat.ML","stat.TH"],"title_canon_sha256":"ad473cb0556a0591887d43d7bee52bbaf6ebd508d943dbcf8ef84b2549626e32","abstract_canon_sha256":"7ae54aedf47f1a28ceb464c30818c4cc57570a5b08b8a198b5f1e218f7f2642e"},"schema_version":"1.0"},"canonical_sha256":"741ad71cca8502a2ba0766c76153f8c50f463bdab8d860c9661126d2494676f8","source":{"kind":"arxiv","id":"2301.11270","version":5},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2301.11270","created_at":"2026-07-05T07:42:38Z"},{"alias_kind":"arxiv_version","alias_value":"2301.11270v5","created_at":"2026-07-05T07:42:38Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2301.11270","created_at":"2026-07-05T07:42:38Z"},{"alias_kind":"pith_short_12","alias_value":"OQNNOHGKQUBK","created_at":"2026-07-05T07:42:38Z"},{"alias_kind":"pith_short_16","alias_value":"OQNNOHGKQUBKFOQH","created_at":"2026-07-05T07:42:38Z"},{"alias_kind":"pith_short_8","alias_value":"OQNNOHGK","created_at":"2026-07-05T07:42:38Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:OQNNOHGKQUBKFOQHM3DWCU7YYU","target":"record","payload":{"canonical_record":{"source":{"id":"2301.11270","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-01-26T18:07:21Z","cross_cats_sorted":["cs.AI","cs.HC","math.ST","stat.ML","stat.TH"],"title_canon_sha256":"ad473cb0556a0591887d43d7bee52bbaf6ebd508d943dbcf8ef84b2549626e32","abstract_canon_sha256":"7ae54aedf47f1a28ceb464c30818c4cc57570a5b08b8a198b5f1e218f7f2642e"},"schema_version":"1.0"},"canonical_sha256":"741ad71cca8502a2ba0766c76153f8c50f463bdab8d860c9661126d2494676f8","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:42:38.654770Z","signature_b64":"wHG1GPUhCDqVKhK1RoinW6GdbyX9yU832ZZNIVY1nGb/hvqTOoNSm0Ixsv9hazoJbO9Q+3V0UPWGGeIvGd4tDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"741ad71cca8502a2ba0766c76153f8c50f463bdab8d860c9661126d2494676f8","last_reissued_at":"2026-07-05T07:42:38.654208Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:42:38.654208Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2301.11270","source_version":5,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T07:42:38Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"O/pdryGDeoYjdDLAXjZ08QfdFzzmgpGUL4wbFYX1P80YmOm8LDHuPgHVab+qdivAVDcZUNf1JoYAtzTAk2rwBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T04:01:26.596274Z"},"content_sha256":"0b00ac00417c19228f5636bc2542c82be7d0cbc0d73d72269acb833ee86b1158","schema_version":"1.0","event_id":"sha256:0b00ac00417c19228f5636bc2542c82be7d0cbc0d73d72269acb833ee86b1158"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:OQNNOHGKQUBKFOQHM3DWCU7YYU","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Principled Reinforcement Learning with Human Feedback from Pairwise or $K$-wise Comparisons","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.HC","math.ST","stat.ML","stat.TH"],"primary_cat":"cs.LG","authors_text":"Banghua Zhu, Jiantao Jiao, Michael I. Jordan","submitted_at":"2023-01-26T18:07:21Z","abstract_excerpt":"We provide a theoretical framework for Reinforcement Learning with Human Feedback (RLHF). Our analysis shows that when the true reward function is linear, the widely used maximum likelihood estimator (MLE) converges under both the Bradley-Terry-Luce (BTL) model and the Plackett-Luce (PL) model. However, we show that when training a policy based on the learned reward model, MLE fails while a pessimistic MLE provides policies with improved performance under certain coverage assumptions. Additionally, we demonstrate that under the PL model, the true MLE and an alternative MLE that splits the $K$-"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2301.11270","kind":"arxiv","version":5},"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/2301.11270/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":"2026-07-05T07:42:38Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"+N4Gw4O/KZ6vVMRVZDlmAJOfVAna2vI/S4ealDvMSNlzILkFFb1KYvu2BaZYMhne+obOxPI/aFkUyZsHk/WwAA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T04:01:26.597348Z"},"content_sha256":"8fb2f7eb250fdecaf37e354b881872de753c508741f310d526305badeeb1b74b","schema_version":"1.0","event_id":"sha256:8fb2f7eb250fdecaf37e354b881872de753c508741f310d526305badeeb1b74b"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/OQNNOHGKQUBKFOQHM3DWCU7YYU/bundle.json","state_url":"https://pith.science/pith/OQNNOHGKQUBKFOQHM3DWCU7YYU/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/OQNNOHGKQUBKFOQHM3DWCU7YYU/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":"2026-08-10T04:01:26Z","links":{"resolver":"https://pith.science/pith/OQNNOHGKQUBKFOQHM3DWCU7YYU","bundle":"https://pith.science/pith/OQNNOHGKQUBKFOQHM3DWCU7YYU/bundle.json","state":"https://pith.science/pith/OQNNOHGKQUBKFOQHM3DWCU7YYU/state.json","well_known_bundle":"https://pith.science/.well-known/pith/OQNNOHGKQUBKFOQHM3DWCU7YYU/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:OQNNOHGKQUBKFOQHM3DWCU7YYU","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":"7ae54aedf47f1a28ceb464c30818c4cc57570a5b08b8a198b5f1e218f7f2642e","cross_cats_sorted":["cs.AI","cs.HC","math.ST","stat.ML","stat.TH"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-01-26T18:07:21Z","title_canon_sha256":"ad473cb0556a0591887d43d7bee52bbaf6ebd508d943dbcf8ef84b2549626e32"},"schema_version":"1.0","source":{"id":"2301.11270","kind":"arxiv","version":5}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2301.11270","created_at":"2026-07-05T07:42:38Z"},{"alias_kind":"arxiv_version","alias_value":"2301.11270v5","created_at":"2026-07-05T07:42:38Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2301.11270","created_at":"2026-07-05T07:42:38Z"},{"alias_kind":"pith_short_12","alias_value":"OQNNOHGKQUBK","created_at":"2026-07-05T07:42:38Z"},{"alias_kind":"pith_short_16","alias_value":"OQNNOHGKQUBKFOQH","created_at":"2026-07-05T07:42:38Z"},{"alias_kind":"pith_short_8","alias_value":"OQNNOHGK","created_at":"2026-07-05T07:42:38Z"}],"graph_snapshots":[{"event_id":"sha256:8fb2f7eb250fdecaf37e354b881872de753c508741f310d526305badeeb1b74b","target":"graph","created_at":"2026-07-05T07:42:38Z","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/2301.11270/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"We provide a theoretical framework for Reinforcement Learning with Human Feedback (RLHF). Our analysis shows that when the true reward function is linear, the widely used maximum likelihood estimator (MLE) converges under both the Bradley-Terry-Luce (BTL) model and the Plackett-Luce (PL) model. However, we show that when training a policy based on the learned reward model, MLE fails while a pessimistic MLE provides policies with improved performance under certain coverage assumptions. Additionally, we demonstrate that under the PL model, the true MLE and an alternative MLE that splits the $K$-","authors_text":"Banghua Zhu, Jiantao Jiao, Michael I. Jordan","cross_cats":["cs.AI","cs.HC","math.ST","stat.ML","stat.TH"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-01-26T18:07:21Z","title":"Principled Reinforcement Learning with Human Feedback from Pairwise or $K$-wise Comparisons"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2301.11270","kind":"arxiv","version":5},"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":"sha256:0b00ac00417c19228f5636bc2542c82be7d0cbc0d73d72269acb833ee86b1158","target":"record","created_at":"2026-07-05T07:42:38Z","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":"7ae54aedf47f1a28ceb464c30818c4cc57570a5b08b8a198b5f1e218f7f2642e","cross_cats_sorted":["cs.AI","cs.HC","math.ST","stat.ML","stat.TH"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-01-26T18:07:21Z","title_canon_sha256":"ad473cb0556a0591887d43d7bee52bbaf6ebd508d943dbcf8ef84b2549626e32"},"schema_version":"1.0","source":{"id":"2301.11270","kind":"arxiv","version":5}},"canonical_sha256":"741ad71cca8502a2ba0766c76153f8c50f463bdab8d860c9661126d2494676f8","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"741ad71cca8502a2ba0766c76153f8c50f463bdab8d860c9661126d2494676f8","first_computed_at":"2026-07-05T07:42:38.654208Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T07:42:38.654208Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"wHG1GPUhCDqVKhK1RoinW6GdbyX9yU832ZZNIVY1nGb/hvqTOoNSm0Ixsv9hazoJbO9Q+3V0UPWGGeIvGd4tDQ==","signature_status":"signed_v1","signed_at":"2026-07-05T07:42:38.654770Z","signed_message":"canonical_sha256_bytes"},"source_id":"2301.11270","source_kind":"arxiv","source_version":5}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:0b00ac00417c19228f5636bc2542c82be7d0cbc0d73d72269acb833ee86b1158","sha256:8fb2f7eb250fdecaf37e354b881872de753c508741f310d526305badeeb1b74b"],"state_sha256":"06982ce5c5d3a9e992689c13321a4ee0e49007fd446b1753774b24d35d9e449a"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Qfxkwb1gJy/R373AwBqtHbVGDbhXFAKVHsK1jOrdVJfMd1Cn4arjeQj3Htadghf1qgUMDJ8YUTg4C1nuulkOCA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-10T04:01:26.604571Z","bundle_sha256":"a88e94f49cec5c610b4cc7400cc07bcfe55eeb9244b36450ffee91f07d6f4058"}}