{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:IHFZODKQQJYE2Z27EINNJ4KXUM","short_pith_number":"pith:IHFZODKQ","canonical_record":{"source":{"id":"2405.19544","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-05-29T22:12:52Z","cross_cats_sorted":["cs.CL","cs.LG","math.OC","stat.ML"],"title_canon_sha256":"d46dd572ec3f5618ebea3d7f6c29db0491f1792d4005755e0fe5010ab4eb2c28","abstract_canon_sha256":"dbd995990b49a402f0ffa513993710bbff46f570d349db22e3c14ca02f001b3d"},"schema_version":"1.0"},"canonical_sha256":"41cb970d5082704d675f221ad4f157a302dc40d48b6e687c6c7bf2816b10746f","source":{"kind":"arxiv","id":"2405.19544","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2405.19544","created_at":"2026-07-05T09:39:06Z"},{"alias_kind":"arxiv_version","alias_value":"2405.19544v3","created_at":"2026-07-05T09:39:06Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.19544","created_at":"2026-07-05T09:39:06Z"},{"alias_kind":"pith_short_12","alias_value":"IHFZODKQQJYE","created_at":"2026-07-05T09:39:06Z"},{"alias_kind":"pith_short_16","alias_value":"IHFZODKQQJYE2Z27","created_at":"2026-07-05T09:39:06Z"},{"alias_kind":"pith_short_8","alias_value":"IHFZODKQ","created_at":"2026-07-05T09:39:06Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:IHFZODKQQJYE2Z27EINNJ4KXUM","target":"record","payload":{"canonical_record":{"source":{"id":"2405.19544","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-05-29T22:12:52Z","cross_cats_sorted":["cs.CL","cs.LG","math.OC","stat.ML"],"title_canon_sha256":"d46dd572ec3f5618ebea3d7f6c29db0491f1792d4005755e0fe5010ab4eb2c28","abstract_canon_sha256":"dbd995990b49a402f0ffa513993710bbff46f570d349db22e3c14ca02f001b3d"},"schema_version":"1.0"},"canonical_sha256":"41cb970d5082704d675f221ad4f157a302dc40d48b6e687c6c7bf2816b10746f","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:39:06.771618Z","signature_b64":"jc+1E7Itk9X2cN9lKyQsRr6f9KuUxVZ6AMGn7OyZK46umgg7vnMqSRKb0u1ZNpaEMYl+9QVhC0kPkWXaU2LkAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"41cb970d5082704d675f221ad4f157a302dc40d48b6e687c6c7bf2816b10746f","last_reissued_at":"2026-07-05T09:39:06.771131Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:39:06.771131Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2405.19544","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":"2026-07-05T09:39:06Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"nZXWKFrmpxrWBc0eA3UFZHsJk8hgEJBxksg84O6c0YHgi+5soxIcCduzO/cbg/Nxo136VjpDVcOP7jzGHPV2Bg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T05:02:42.078231Z"},"content_sha256":"8c848cc058cb97aece89221103bf7d046c13fe5e054ba3b9d3554916dc0a68d4","schema_version":"1.0","event_id":"sha256:8c848cc058cb97aece89221103bf7d046c13fe5e054ba3b9d3554916dc0a68d4"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:IHFZODKQQJYE2Z27EINNJ4KXUM","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"One-Shot Safety Alignment for Large Language Models via Optimal Dualization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.LG","math.OC","stat.ML"],"primary_cat":"cs.AI","authors_text":"Dongsheng Ding, Edgar Dobriban, Hamed Hassani, Osbert Bastani, Shuo Li, Xinmeng Huang","submitted_at":"2024-05-29T22:12:52Z","abstract_excerpt":"The growing safety concerns surrounding large language models raise an urgent need to align them with diverse human preferences to simultaneously enhance their helpfulness and safety. A promising approach is to enforce safety constraints through Reinforcement Learning from Human Feedback (RLHF). For such constrained RLHF, typical Lagrangian-based primal-dual policy optimization methods are computationally expensive and often unstable. This paper presents a perspective of dualization that reduces constrained alignment to an equivalent unconstrained alignment problem. We do so by pre-optimizing "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.19544","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/2405.19544/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":"2026-07-05T09:39:06Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"A7hxZUo2FT8XcMIb/FpZDseKK5uGYGS1KueDlBLNLGo3K6Q4IMXqV4xzOxz36MD8VAAhQC688TKSnpFESDj0Cg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T05:02:42.079263Z"},"content_sha256":"6e6ae0ec5462eb9724063d3dd3a031b6f4e36d0ad0c680209ec6841b9d83cd2d","schema_version":"1.0","event_id":"sha256:6e6ae0ec5462eb9724063d3dd3a031b6f4e36d0ad0c680209ec6841b9d83cd2d"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/IHFZODKQQJYE2Z27EINNJ4KXUM/bundle.json","state_url":"https://pith.science/pith/IHFZODKQQJYE2Z27EINNJ4KXUM/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/IHFZODKQQJYE2Z27EINNJ4KXUM/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":"2026-08-10T05:02:42Z","links":{"resolver":"https://pith.science/pith/IHFZODKQQJYE2Z27EINNJ4KXUM","bundle":"https://pith.science/pith/IHFZODKQQJYE2Z27EINNJ4KXUM/bundle.json","state":"https://pith.science/pith/IHFZODKQQJYE2Z27EINNJ4KXUM/state.json","well_known_bundle":"https://pith.science/.well-known/pith/IHFZODKQQJYE2Z27EINNJ4KXUM/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:IHFZODKQQJYE2Z27EINNJ4KXUM","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":"dbd995990b49a402f0ffa513993710bbff46f570d349db22e3c14ca02f001b3d","cross_cats_sorted":["cs.CL","cs.LG","math.OC","stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-05-29T22:12:52Z","title_canon_sha256":"d46dd572ec3f5618ebea3d7f6c29db0491f1792d4005755e0fe5010ab4eb2c28"},"schema_version":"1.0","source":{"id":"2405.19544","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2405.19544","created_at":"2026-07-05T09:39:06Z"},{"alias_kind":"arxiv_version","alias_value":"2405.19544v3","created_at":"2026-07-05T09:39:06Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.19544","created_at":"2026-07-05T09:39:06Z"},{"alias_kind":"pith_short_12","alias_value":"IHFZODKQQJYE","created_at":"2026-07-05T09:39:06Z"},{"alias_kind":"pith_short_16","alias_value":"IHFZODKQQJYE2Z27","created_at":"2026-07-05T09:39:06Z"},{"alias_kind":"pith_short_8","alias_value":"IHFZODKQ","created_at":"2026-07-05T09:39:06Z"}],"graph_snapshots":[{"event_id":"sha256:6e6ae0ec5462eb9724063d3dd3a031b6f4e36d0ad0c680209ec6841b9d83cd2d","target":"graph","created_at":"2026-07-05T09:39:06Z","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/2405.19544/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"The growing safety concerns surrounding large language models raise an urgent need to align them with diverse human preferences to simultaneously enhance their helpfulness and safety. A promising approach is to enforce safety constraints through Reinforcement Learning from Human Feedback (RLHF). For such constrained RLHF, typical Lagrangian-based primal-dual policy optimization methods are computationally expensive and often unstable. This paper presents a perspective of dualization that reduces constrained alignment to an equivalent unconstrained alignment problem. We do so by pre-optimizing ","authors_text":"Dongsheng Ding, Edgar Dobriban, Hamed Hassani, Osbert Bastani, Shuo Li, Xinmeng Huang","cross_cats":["cs.CL","cs.LG","math.OC","stat.ML"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-05-29T22:12:52Z","title":"One-Shot Safety Alignment for Large Language Models via Optimal Dualization"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.19544","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":"sha256:8c848cc058cb97aece89221103bf7d046c13fe5e054ba3b9d3554916dc0a68d4","target":"record","created_at":"2026-07-05T09:39:06Z","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":"dbd995990b49a402f0ffa513993710bbff46f570d349db22e3c14ca02f001b3d","cross_cats_sorted":["cs.CL","cs.LG","math.OC","stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-05-29T22:12:52Z","title_canon_sha256":"d46dd572ec3f5618ebea3d7f6c29db0491f1792d4005755e0fe5010ab4eb2c28"},"schema_version":"1.0","source":{"id":"2405.19544","kind":"arxiv","version":3}},"canonical_sha256":"41cb970d5082704d675f221ad4f157a302dc40d48b6e687c6c7bf2816b10746f","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"41cb970d5082704d675f221ad4f157a302dc40d48b6e687c6c7bf2816b10746f","first_computed_at":"2026-07-05T09:39:06.771131Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T09:39:06.771131Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"jc+1E7Itk9X2cN9lKyQsRr6f9KuUxVZ6AMGn7OyZK46umgg7vnMqSRKb0u1ZNpaEMYl+9QVhC0kPkWXaU2LkAQ==","signature_status":"signed_v1","signed_at":"2026-07-05T09:39:06.771618Z","signed_message":"canonical_sha256_bytes"},"source_id":"2405.19544","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:8c848cc058cb97aece89221103bf7d046c13fe5e054ba3b9d3554916dc0a68d4","sha256:6e6ae0ec5462eb9724063d3dd3a031b6f4e36d0ad0c680209ec6841b9d83cd2d"],"state_sha256":"9fd149b5f1aaeadd200242805a095c99c07203e2f569d95fe93a19813aa6fc6f"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"zQx+vTG8L+zglCN4fx/O4hiTYKzc58EvcxgIiohqu+tS6QXlfMF/DpysEOA97P/OP0sRmyxftupyMSVNuCrOCA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-10T05:02:42.084923Z","bundle_sha256":"318d1f2f1ed2e7b2792cd5301b6d8c8a5fdcd387c877b05efeebd52a9d914e64"}}