{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:EIUNGFN63U756SOQVID6C7AI6X","short_pith_number":"pith:EIUNGFN6","canonical_record":{"source":{"id":"2503.18130","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-03-23T16:20:59Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"a63c3e94f31705cb0ab702c2bd7b63449edcd63dc40f5f703f618cda8f372b34","abstract_canon_sha256":"7ea2787c4e7da787feb91acea81b32f5c5264898cdefbff30c0ec13b30474a16"},"schema_version":"1.0"},"canonical_sha256":"2228d315bedd3fdf49d0aa07e17c08f5e0310c70c2d47d886e29cac8498f4544","source":{"kind":"arxiv","id":"2503.18130","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2503.18130","created_at":"2026-07-05T10:38:03Z"},{"alias_kind":"arxiv_version","alias_value":"2503.18130v1","created_at":"2026-07-05T10:38:03Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.18130","created_at":"2026-07-05T10:38:03Z"},{"alias_kind":"pith_short_12","alias_value":"EIUNGFN63U75","created_at":"2026-07-05T10:38:03Z"},{"alias_kind":"pith_short_16","alias_value":"EIUNGFN63U756SOQ","created_at":"2026-07-05T10:38:03Z"},{"alias_kind":"pith_short_8","alias_value":"EIUNGFN6","created_at":"2026-07-05T10:38:03Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:EIUNGFN63U756SOQVID6C7AI6X","target":"record","payload":{"canonical_record":{"source":{"id":"2503.18130","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-03-23T16:20:59Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"a63c3e94f31705cb0ab702c2bd7b63449edcd63dc40f5f703f618cda8f372b34","abstract_canon_sha256":"7ea2787c4e7da787feb91acea81b32f5c5264898cdefbff30c0ec13b30474a16"},"schema_version":"1.0"},"canonical_sha256":"2228d315bedd3fdf49d0aa07e17c08f5e0310c70c2d47d886e29cac8498f4544","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:38:03.414569Z","signature_b64":"JMQu3xBkPghkN1BwQQMcKM39poIL6VowuiEIwJGOkKIb4kamswjTfujwXMBvo8PaRAO73z+fu5TlKURFFrBJCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2228d315bedd3fdf49d0aa07e17c08f5e0310c70c2d47d886e29cac8498f4544","last_reissued_at":"2026-07-05T10:38:03.414122Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:38:03.414122Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2503.18130","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:38:03Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"8DvC10B6uXyw5k4SWwoxsO4TXsrSJtBLactqm8pKjykBMjpVtAAr8y/F7coIye4orbz+fOy7Q5ubJcOk9BFaBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T10:00:28.821746Z"},"content_sha256":"421c71cd2e08f34bd68ba2906991c5d25fa546bd38c9a2cad7aab19038e692f3","schema_version":"1.0","event_id":"sha256:421c71cd2e08f34bd68ba2906991c5d25fa546bd38c9a2cad7aab19038e692f3"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:EIUNGFN63U756SOQVID6C7AI6X","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Mitigating Reward Over-Optimization in RLHF via Behavior-Supported Regularization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Gang Pan, Juntao Dai, Qian Zheng, Taiye Chen, Yaodong Yang","submitted_at":"2025-03-23T16:20:59Z","abstract_excerpt":"Reinforcement learning from human feedback (RLHF) is an effective method for aligning large language models (LLMs) with human values. However, reward over-optimization remains an open challenge leading to discrepancies between the performance of LLMs under the reward model and the true human objectives. A primary contributor to reward over-optimization is the extrapolation error that arises when the reward model evaluates out-of-distribution (OOD) responses. However, current methods still fail to prevent the increasing frequency of OOD response generation during the reinforcement learning (RL)"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.18130","kind":"arxiv","version":1},"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/2503.18130/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":"2026-07-05T10:38:03Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"JzNQ9KvICgfE5Zxviltk7RrBnsTmtbjeJ8WfQDugJghFXrjl67ZAAPAQEkUEV0KvYWs3wLdZnuRSRRyu748gBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T10:00:28.822232Z"},"content_sha256":"8b56ff97e4b9c2e1aeff609035882bf8576f286e251ecabd5bc6f1b2da26ce95","schema_version":"1.0","event_id":"sha256:8b56ff97e4b9c2e1aeff609035882bf8576f286e251ecabd5bc6f1b2da26ce95"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/EIUNGFN63U756SOQVID6C7AI6X/bundle.json","state_url":"https://pith.science/pith/EIUNGFN63U756SOQVID6C7AI6X/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/EIUNGFN63U756SOQVID6C7AI6X/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":"2026-08-09T10:00:28Z","links":{"resolver":"https://pith.science/pith/EIUNGFN63U756SOQVID6C7AI6X","bundle":"https://pith.science/pith/EIUNGFN63U756SOQVID6C7AI6X/bundle.json","state":"https://pith.science/pith/EIUNGFN63U756SOQVID6C7AI6X/state.json","well_known_bundle":"https://pith.science/.well-known/pith/EIUNGFN63U756SOQVID6C7AI6X/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:EIUNGFN63U756SOQVID6C7AI6X","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":"7ea2787c4e7da787feb91acea81b32f5c5264898cdefbff30c0ec13b30474a16","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-03-23T16:20:59Z","title_canon_sha256":"a63c3e94f31705cb0ab702c2bd7b63449edcd63dc40f5f703f618cda8f372b34"},"schema_version":"1.0","source":{"id":"2503.18130","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2503.18130","created_at":"2026-07-05T10:38:03Z"},{"alias_kind":"arxiv_version","alias_value":"2503.18130v1","created_at":"2026-07-05T10:38:03Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.18130","created_at":"2026-07-05T10:38:03Z"},{"alias_kind":"pith_short_12","alias_value":"EIUNGFN63U75","created_at":"2026-07-05T10:38:03Z"},{"alias_kind":"pith_short_16","alias_value":"EIUNGFN63U756SOQ","created_at":"2026-07-05T10:38:03Z"},{"alias_kind":"pith_short_8","alias_value":"EIUNGFN6","created_at":"2026-07-05T10:38:03Z"}],"graph_snapshots":[{"event_id":"sha256:8b56ff97e4b9c2e1aeff609035882bf8576f286e251ecabd5bc6f1b2da26ce95","target":"graph","created_at":"2026-07-05T10:38:03Z","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/2503.18130/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reinforcement learning from human feedback (RLHF) is an effective method for aligning large language models (LLMs) with human values. However, reward over-optimization remains an open challenge leading to discrepancies between the performance of LLMs under the reward model and the true human objectives. A primary contributor to reward over-optimization is the extrapolation error that arises when the reward model evaluates out-of-distribution (OOD) responses. However, current methods still fail to prevent the increasing frequency of OOD response generation during the reinforcement learning (RL)","authors_text":"Gang Pan, Juntao Dai, Qian Zheng, Taiye Chen, Yaodong Yang","cross_cats":["cs.AI"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-03-23T16:20:59Z","title":"Mitigating Reward Over-Optimization in RLHF via Behavior-Supported Regularization"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.18130","kind":"arxiv","version":1},"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":"sha256:421c71cd2e08f34bd68ba2906991c5d25fa546bd38c9a2cad7aab19038e692f3","target":"record","created_at":"2026-07-05T10:38:03Z","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":"7ea2787c4e7da787feb91acea81b32f5c5264898cdefbff30c0ec13b30474a16","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-03-23T16:20:59Z","title_canon_sha256":"a63c3e94f31705cb0ab702c2bd7b63449edcd63dc40f5f703f618cda8f372b34"},"schema_version":"1.0","source":{"id":"2503.18130","kind":"arxiv","version":1}},"canonical_sha256":"2228d315bedd3fdf49d0aa07e17c08f5e0310c70c2d47d886e29cac8498f4544","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"2228d315bedd3fdf49d0aa07e17c08f5e0310c70c2d47d886e29cac8498f4544","first_computed_at":"2026-07-05T10:38:03.414122Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:38:03.414122Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"JMQu3xBkPghkN1BwQQMcKM39poIL6VowuiEIwJGOkKIb4kamswjTfujwXMBvo8PaRAO73z+fu5TlKURFFrBJCw==","signature_status":"signed_v1","signed_at":"2026-07-05T10:38:03.414569Z","signed_message":"canonical_sha256_bytes"},"source_id":"2503.18130","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:421c71cd2e08f34bd68ba2906991c5d25fa546bd38c9a2cad7aab19038e692f3","sha256:8b56ff97e4b9c2e1aeff609035882bf8576f286e251ecabd5bc6f1b2da26ce95"],"state_sha256":"d83023307d7c5373f8f1d0312d24d8a86399629c706f59e6ebd4b1d4d6cc0057"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"lYGvFX4xBHmmoG4BFFx0ky5oCjOYOl/06KICpPiuuMaBhp8aoCBCbuijiEO3fqZ1y2roIZzv8vC2XAx07iZsCA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-09T10:00:28.825694Z","bundle_sha256":"ac5ebe1b1e8e356932fa365fdf950427c43b1049a0039f891af4c984c807218f"}}