{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:JPV2REKRDDKEMZ5ALECYX6O33I","short_pith_number":"pith:JPV2REKR","canonical_record":{"source":{"id":"2501.08617","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-01-15T06:33:15Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"4f112f0207ec733cb1f093c7b01a6fb3fa7358a5850c74055ceb60d46752083c","abstract_canon_sha256":"df0f822d09330ebe0ce2cd13d1b466abbfda52bf14947064f702a32bb9136caf"},"schema_version":"1.0"},"canonical_sha256":"4beba8915118d44667a059058bf9dbda1ec9dc3efc01ef1ef87db701e3b26ef8","source":{"kind":"arxiv","id":"2501.08617","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2501.08617","created_at":"2026-07-05T11:18:41Z"},{"alias_kind":"arxiv_version","alias_value":"2501.08617v3","created_at":"2026-07-05T11:18:41Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.08617","created_at":"2026-07-05T11:18:41Z"},{"alias_kind":"pith_short_12","alias_value":"JPV2REKRDDKE","created_at":"2026-07-05T11:18:41Z"},{"alias_kind":"pith_short_16","alias_value":"JPV2REKRDDKEMZ5A","created_at":"2026-07-05T11:18:41Z"},{"alias_kind":"pith_short_8","alias_value":"JPV2REKR","created_at":"2026-07-05T11:18:41Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:JPV2REKRDDKEMZ5ALECYX6O33I","target":"record","payload":{"canonical_record":{"source":{"id":"2501.08617","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-01-15T06:33:15Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"4f112f0207ec733cb1f093c7b01a6fb3fa7358a5850c74055ceb60d46752083c","abstract_canon_sha256":"df0f822d09330ebe0ce2cd13d1b466abbfda52bf14947064f702a32bb9136caf"},"schema_version":"1.0"},"canonical_sha256":"4beba8915118d44667a059058bf9dbda1ec9dc3efc01ef1ef87db701e3b26ef8","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:18:41.171772Z","signature_b64":"Cc0SsPsAk3PkPXH3pcFQFwnB+1FqJeK3EkbtQj+DBnRwsREP9J+e1C5Xkla22+AkMD98vPN+10Xs89hQOGk/AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4beba8915118d44667a059058bf9dbda1ec9dc3efc01ef1ef87db701e3b26ef8","last_reissued_at":"2026-07-05T11:18:41.171327Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:18:41.171327Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2501.08617","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:18:41Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"+4iz4dL5CiMkTlz3x99MRl4AJbphHwneBaIX8q1cSqVV2eCB+q2ctrjR0HQxTU+F9bXjjsEKv1VNk//1YcMgDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T12:39:41.900539Z"},"content_sha256":"5bfa2abadf3bc91aedf53e8340037bd8b68966e6ba9c18691dced4fe3b9f1f1a","schema_version":"1.0","event_id":"sha256:5bfa2abadf3bc91aedf53e8340037bd8b68966e6ba9c18691dced4fe3b9f1f1a"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:JPV2REKRDDKEMZ5ALECYX6O33I","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"RLHS: Mitigating Misalignment in RLHF with Hindsight Simulation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Haimin Hu, Jaime Fern\\'andez Fisac, Kaiqu Liang, Ryan Liu, Thomas L. Griffiths","submitted_at":"2025-01-15T06:33:15Z","abstract_excerpt":"While Reinforcement Learning from Human Feedback (RLHF) has shown promise in aligning generative AI, we present empirical evidence that it can also cause severe, systematic misalignment. We hypothesize that this stems from evaluator feedback depending on downstream outcome predictions (foresight) that can be influenced by the AI's output, inducing Goodhart's law dynamics. We present a theoretical analysis showing that conditioning evaluator feedback on downstream observations (hindsight) inhibits this effect by decoupling the alignment signal from potentially compromised predictions--crucially"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.08617","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/2501.08617/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":"2026-07-05T11:18:41Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"76Y+XNp4WTNLD2lR5XTTPf3tP+YgTXNyj1Hhh7CXoV3YySZy4EcU2J9THrnVJDEGPcUYUSuXHJDDyGHRJihzCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T12:39:41.901312Z"},"content_sha256":"4f0ada92bc3181605fa53f08d9b2ab9fba09815ecd3e85109158038f7a9f5344","schema_version":"1.0","event_id":"sha256:4f0ada92bc3181605fa53f08d9b2ab9fba09815ecd3e85109158038f7a9f5344"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/JPV2REKRDDKEMZ5ALECYX6O33I/bundle.json","state_url":"https://pith.science/pith/JPV2REKRDDKEMZ5ALECYX6O33I/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/JPV2REKRDDKEMZ5ALECYX6O33I/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":"2026-08-10T12:39:41Z","links":{"resolver":"https://pith.science/pith/JPV2REKRDDKEMZ5ALECYX6O33I","bundle":"https://pith.science/pith/JPV2REKRDDKEMZ5ALECYX6O33I/bundle.json","state":"https://pith.science/pith/JPV2REKRDDKEMZ5ALECYX6O33I/state.json","well_known_bundle":"https://pith.science/.well-known/pith/JPV2REKRDDKEMZ5ALECYX6O33I/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:JPV2REKRDDKEMZ5ALECYX6O33I","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":"df0f822d09330ebe0ce2cd13d1b466abbfda52bf14947064f702a32bb9136caf","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-01-15T06:33:15Z","title_canon_sha256":"4f112f0207ec733cb1f093c7b01a6fb3fa7358a5850c74055ceb60d46752083c"},"schema_version":"1.0","source":{"id":"2501.08617","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2501.08617","created_at":"2026-07-05T11:18:41Z"},{"alias_kind":"arxiv_version","alias_value":"2501.08617v3","created_at":"2026-07-05T11:18:41Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.08617","created_at":"2026-07-05T11:18:41Z"},{"alias_kind":"pith_short_12","alias_value":"JPV2REKRDDKE","created_at":"2026-07-05T11:18:41Z"},{"alias_kind":"pith_short_16","alias_value":"JPV2REKRDDKEMZ5A","created_at":"2026-07-05T11:18:41Z"},{"alias_kind":"pith_short_8","alias_value":"JPV2REKR","created_at":"2026-07-05T11:18:41Z"}],"graph_snapshots":[{"event_id":"sha256:4f0ada92bc3181605fa53f08d9b2ab9fba09815ecd3e85109158038f7a9f5344","target":"graph","created_at":"2026-07-05T11:18:41Z","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/2501.08617/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"While Reinforcement Learning from Human Feedback (RLHF) has shown promise in aligning generative AI, we present empirical evidence that it can also cause severe, systematic misalignment. We hypothesize that this stems from evaluator feedback depending on downstream outcome predictions (foresight) that can be influenced by the AI's output, inducing Goodhart's law dynamics. We present a theoretical analysis showing that conditioning evaluator feedback on downstream observations (hindsight) inhibits this effect by decoupling the alignment signal from potentially compromised predictions--crucially","authors_text":"Haimin Hu, Jaime Fern\\'andez Fisac, Kaiqu Liang, Ryan Liu, Thomas L. Griffiths","cross_cats":["cs.AI","cs.CL"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-01-15T06:33:15Z","title":"RLHS: Mitigating Misalignment in RLHF with Hindsight Simulation"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.08617","kind":"arxiv","version":3},"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":"sha256:5bfa2abadf3bc91aedf53e8340037bd8b68966e6ba9c18691dced4fe3b9f1f1a","target":"record","created_at":"2026-07-05T11:18:41Z","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":"df0f822d09330ebe0ce2cd13d1b466abbfda52bf14947064f702a32bb9136caf","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-01-15T06:33:15Z","title_canon_sha256":"4f112f0207ec733cb1f093c7b01a6fb3fa7358a5850c74055ceb60d46752083c"},"schema_version":"1.0","source":{"id":"2501.08617","kind":"arxiv","version":3}},"canonical_sha256":"4beba8915118d44667a059058bf9dbda1ec9dc3efc01ef1ef87db701e3b26ef8","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"4beba8915118d44667a059058bf9dbda1ec9dc3efc01ef1ef87db701e3b26ef8","first_computed_at":"2026-07-05T11:18:41.171327Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:18:41.171327Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"Cc0SsPsAk3PkPXH3pcFQFwnB+1FqJeK3EkbtQj+DBnRwsREP9J+e1C5Xkla22+AkMD98vPN+10Xs89hQOGk/AQ==","signature_status":"signed_v1","signed_at":"2026-07-05T11:18:41.171772Z","signed_message":"canonical_sha256_bytes"},"source_id":"2501.08617","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:5bfa2abadf3bc91aedf53e8340037bd8b68966e6ba9c18691dced4fe3b9f1f1a","sha256:4f0ada92bc3181605fa53f08d9b2ab9fba09815ecd3e85109158038f7a9f5344"],"state_sha256":"42b4e947d9d5ba2818c7d14595a81797af1c777470a12fb01e16223558afea31"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6npW3p1rtn/Lo5FIVw2YWMp2FHp3hzm3TjsSeEhRkQxt26+AYiJgHha/Zb8InINcJq614tr7UE3s5wzQR/+5Ag==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-10T12:39:41.919482Z","bundle_sha256":"686f3ff827520f6ab2b09e12c7794a44ba31c9fe667c7d1ec7d33f3d13329633"}}