{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:E3VDQ5S76MEE4A3DWQYTNEJ6SH","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":"c4fdfde24d51eebe4b1ede64a8727e934e1f527bb3c2335b48637dc3e486a10a","cross_cats_sorted":["cs.AI","cs.CL","stat.AP"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-09T22:03:56Z","title_canon_sha256":"06693bab1075ada37e93ef2686655ea9fd294f6722299c9f2edd28c3cc94a690"},"schema_version":"1.0","source":{"id":"2506.08266","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2506.08266","created_at":"2026-07-05T11:18:47Z"},{"alias_kind":"arxiv_version","alias_value":"2506.08266v1","created_at":"2026-07-05T11:18:47Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.08266","created_at":"2026-07-05T11:18:47Z"},{"alias_kind":"pith_short_12","alias_value":"E3VDQ5S76MEE","created_at":"2026-07-05T11:18:47Z"},{"alias_kind":"pith_short_16","alias_value":"E3VDQ5S76MEE4A3D","created_at":"2026-07-05T11:18:47Z"},{"alias_kind":"pith_short_8","alias_value":"E3VDQ5S7","created_at":"2026-07-05T11:18:47Z"}],"graph_snapshots":[{"event_id":"sha256:3a1b54b04e72b385e7af5477769778b481d2f5d6760b1d286dc66bdd5c11a325","target":"graph","created_at":"2026-07-05T11:18:47Z","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/2506.08266/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Existing approaches to language model alignment often treat safety as a tradeoff against helpfulness, which can lead to unacceptable responses in sensitive domains. To ensure reliable performance in such settings, we propose High-Confidence Safe Reinforcement Learning from Human Feedback (HC-RLHF), a method that provides high-confidence safety guarantees while maximizing helpfulness. Similar to previous methods, HC-RLHF explicitly decouples human preferences into helpfulness and harmlessness (safety), which are learned by training a reward model and a cost model, respectively. It then employs ","authors_text":"Austin Hoag, Blossom Metevier, Philip S. Thomas, Scott Niekum, Will Schwarzer, Yaswanth Chittepu","cross_cats":["cs.AI","cs.CL","stat.AP"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-09T22:03:56Z","title":"Reinforcement Learning from Human Feedback with High-Confidence Safety Constraints"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.08266","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":"sha256:60632c54b349a4aeffa564311c64e41707d905115a62e9241caaccc9e0cfbffd","target":"record","created_at":"2026-07-05T11:18:47Z","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":"c4fdfde24d51eebe4b1ede64a8727e934e1f527bb3c2335b48637dc3e486a10a","cross_cats_sorted":["cs.AI","cs.CL","stat.AP"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-09T22:03:56Z","title_canon_sha256":"06693bab1075ada37e93ef2686655ea9fd294f6722299c9f2edd28c3cc94a690"},"schema_version":"1.0","source":{"id":"2506.08266","kind":"arxiv","version":1}},"canonical_sha256":"26ea38765ff3084e0363b43136913e91e7e9e6866df4e211a0f9421dac716ff2","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"26ea38765ff3084e0363b43136913e91e7e9e6866df4e211a0f9421dac716ff2","first_computed_at":"2026-07-05T11:18:47.184938Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:18:47.184938Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"dU1RqoPudM3KHKeAGc2ZI2op2gMO0+iX/iF2Mmn21s6/dgv/hdTR/tZUgXb/aM6KQHwCWCRAICFU6Jk4bZ0oDQ==","signature_status":"signed_v1","signed_at":"2026-07-05T11:18:47.185421Z","signed_message":"canonical_sha256_bytes"},"source_id":"2506.08266","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:60632c54b349a4aeffa564311c64e41707d905115a62e9241caaccc9e0cfbffd","sha256:3a1b54b04e72b385e7af5477769778b481d2f5d6760b1d286dc66bdd5c11a325"],"state_sha256":"2da9d88ed23dadb7f6ecacca647745f0a3da984974c3075d489d3f85e78e0cb9"}