{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:5LY6MINNYDJJRNGJ6VPSHEOPI6","short_pith_number":"pith:5LY6MINN","canonical_record":{"source":{"id":"2408.15313","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-08-27T17:31:21Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"8d0706387fae7edc0dc8e066462d213906ee3be0f694670f0a28e0632354fc9f","abstract_canon_sha256":"c4ce0faa4be9b7a2706c2bfdc73968fababd1ba7a99e888e85d7a9eccf3e0dbd"},"schema_version":"1.0"},"canonical_sha256":"eaf1e621adc0d298b4c9f55f2391cf47a486bea5c4cdaca11477e14bc52a87a5","source":{"kind":"arxiv","id":"2408.15313","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2408.15313","created_at":"2026-07-05T10:45:54Z"},{"alias_kind":"arxiv_version","alias_value":"2408.15313v2","created_at":"2026-07-05T10:45:54Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.15313","created_at":"2026-07-05T10:45:54Z"},{"alias_kind":"pith_short_12","alias_value":"5LY6MINNYDJJ","created_at":"2026-07-05T10:45:54Z"},{"alias_kind":"pith_short_16","alias_value":"5LY6MINNYDJJRNGJ","created_at":"2026-07-05T10:45:54Z"},{"alias_kind":"pith_short_8","alias_value":"5LY6MINN","created_at":"2026-07-05T10:45:54Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:5LY6MINNYDJJRNGJ6VPSHEOPI6","target":"record","payload":{"canonical_record":{"source":{"id":"2408.15313","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-08-27T17:31:21Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"8d0706387fae7edc0dc8e066462d213906ee3be0f694670f0a28e0632354fc9f","abstract_canon_sha256":"c4ce0faa4be9b7a2706c2bfdc73968fababd1ba7a99e888e85d7a9eccf3e0dbd"},"schema_version":"1.0"},"canonical_sha256":"eaf1e621adc0d298b4c9f55f2391cf47a486bea5c4cdaca11477e14bc52a87a5","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:45:54.696285Z","signature_b64":"FAs45iT0BsTHz12K+cznufewTLcElKuefkNldRO4l1Kymew6oNGbT/HheUIOefy2shBzvVguuJgLNZT+VMYMDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"eaf1e621adc0d298b4c9f55f2391cf47a486bea5c4cdaca11477e14bc52a87a5","last_reissued_at":"2026-07-05T10:45:54.695855Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:45:54.695855Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2408.15313","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:45:54Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"9iWAlKlNpjRp79R7v9GiheU+4oj8EfnPfnx6gJtDoUOUUpTj10UD4MwIJDKy9XyJen+teJ08A1F9CwQMl+O8AA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-22T08:37:07.024550Z"},"content_sha256":"11509deac6b78d64b10b0a68d0c7c47c60880294d16eebc11f31f8a8c310d811","schema_version":"1.0","event_id":"sha256:11509deac6b78d64b10b0a68d0c7c47c60880294d16eebc11f31f8a8c310d811"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:5LY6MINNYDJJRNGJ6VPSHEOPI6","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Bi-Factorial Preference Optimization: Balancing Safety-Helpfulness in Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.AI","authors_text":"Adel Bibi, Mohamed Elhoseiny, Philip H.S. Torr, Wenxuan Zhang","submitted_at":"2024-08-27T17:31:21Z","abstract_excerpt":"Fine-tuning large language models (LLMs) on human preferences, typically through reinforcement learning from human feedback (RLHF), has proven successful in enhancing their capabilities. However, ensuring the safety of LLMs during fine-tuning remains a critical concern, and mitigating the potential conflicts in safety and helpfulness is costly in RLHF. To address this issue, we propose a supervised learning framework called Bi-Factorial Preference Optimization (BFPO), which re-parameterizes a joint RLHF objective of both safety and helpfulness into a single supervised learning objective. In su"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2408.15313","kind":"arxiv","version":2},"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/2408.15313/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":"2026-07-05T10:45:54Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"St8VCxr/0skXCAGEWCefUQAI/mms0ETqZcUPkNRYcVFjauL9P//xwjfC/9t69594V/3mrvs63WoKk7Chf9ehAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-22T08:37:07.025065Z"},"content_sha256":"0c2b4d4f3ff6aa610fdeca7c21c17963ec44ecbc26c7072b7f79d9350a204b21","schema_version":"1.0","event_id":"sha256:0c2b4d4f3ff6aa610fdeca7c21c17963ec44ecbc26c7072b7f79d9350a204b21"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/5LY6MINNYDJJRNGJ6VPSHEOPI6/bundle.json","state_url":"https://pith.science/pith/5LY6MINNYDJJRNGJ6VPSHEOPI6/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/5LY6MINNYDJJRNGJ6VPSHEOPI6/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":"2026-08-22T08:37:07Z","links":{"resolver":"https://pith.science/pith/5LY6MINNYDJJRNGJ6VPSHEOPI6","bundle":"https://pith.science/pith/5LY6MINNYDJJRNGJ6VPSHEOPI6/bundle.json","state":"https://pith.science/pith/5LY6MINNYDJJRNGJ6VPSHEOPI6/state.json","well_known_bundle":"https://pith.science/.well-known/pith/5LY6MINNYDJJRNGJ6VPSHEOPI6/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:5LY6MINNYDJJRNGJ6VPSHEOPI6","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":"c4ce0faa4be9b7a2706c2bfdc73968fababd1ba7a99e888e85d7a9eccf3e0dbd","cross_cats_sorted":["cs.CL","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-08-27T17:31:21Z","title_canon_sha256":"8d0706387fae7edc0dc8e066462d213906ee3be0f694670f0a28e0632354fc9f"},"schema_version":"1.0","source":{"id":"2408.15313","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2408.15313","created_at":"2026-07-05T10:45:54Z"},{"alias_kind":"arxiv_version","alias_value":"2408.15313v2","created_at":"2026-07-05T10:45:54Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.15313","created_at":"2026-07-05T10:45:54Z"},{"alias_kind":"pith_short_12","alias_value":"5LY6MINNYDJJ","created_at":"2026-07-05T10:45:54Z"},{"alias_kind":"pith_short_16","alias_value":"5LY6MINNYDJJRNGJ","created_at":"2026-07-05T10:45:54Z"},{"alias_kind":"pith_short_8","alias_value":"5LY6MINN","created_at":"2026-07-05T10:45:54Z"}],"graph_snapshots":[{"event_id":"sha256:0c2b4d4f3ff6aa610fdeca7c21c17963ec44ecbc26c7072b7f79d9350a204b21","target":"graph","created_at":"2026-07-05T10:45:54Z","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/2408.15313/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Fine-tuning large language models (LLMs) on human preferences, typically through reinforcement learning from human feedback (RLHF), has proven successful in enhancing their capabilities. However, ensuring the safety of LLMs during fine-tuning remains a critical concern, and mitigating the potential conflicts in safety and helpfulness is costly in RLHF. To address this issue, we propose a supervised learning framework called Bi-Factorial Preference Optimization (BFPO), which re-parameterizes a joint RLHF objective of both safety and helpfulness into a single supervised learning objective. In su","authors_text":"Adel Bibi, Mohamed Elhoseiny, Philip H.S. Torr, Wenxuan Zhang","cross_cats":["cs.CL","cs.LG"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-08-27T17:31:21Z","title":"Bi-Factorial Preference Optimization: Balancing Safety-Helpfulness in Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2408.15313","kind":"arxiv","version":2},"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":"sha256:11509deac6b78d64b10b0a68d0c7c47c60880294d16eebc11f31f8a8c310d811","target":"record","created_at":"2026-07-05T10:45:54Z","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":"c4ce0faa4be9b7a2706c2bfdc73968fababd1ba7a99e888e85d7a9eccf3e0dbd","cross_cats_sorted":["cs.CL","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-08-27T17:31:21Z","title_canon_sha256":"8d0706387fae7edc0dc8e066462d213906ee3be0f694670f0a28e0632354fc9f"},"schema_version":"1.0","source":{"id":"2408.15313","kind":"arxiv","version":2}},"canonical_sha256":"eaf1e621adc0d298b4c9f55f2391cf47a486bea5c4cdaca11477e14bc52a87a5","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"eaf1e621adc0d298b4c9f55f2391cf47a486bea5c4cdaca11477e14bc52a87a5","first_computed_at":"2026-07-05T10:45:54.695855Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:45:54.695855Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"FAs45iT0BsTHz12K+cznufewTLcElKuefkNldRO4l1Kymew6oNGbT/HheUIOefy2shBzvVguuJgLNZT+VMYMDQ==","signature_status":"signed_v1","signed_at":"2026-07-05T10:45:54.696285Z","signed_message":"canonical_sha256_bytes"},"source_id":"2408.15313","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:11509deac6b78d64b10b0a68d0c7c47c60880294d16eebc11f31f8a8c310d811","sha256:0c2b4d4f3ff6aa610fdeca7c21c17963ec44ecbc26c7072b7f79d9350a204b21"],"state_sha256":"e37924c240e67871f526f12c3ac6566b73d0c56401cceb37a0faf8d80bcf99c4"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"HRvbZffam2BNQ7huqBIWnkf5kWcwS66/UTVXj9xfvIUMR8W3gLD+Etk7SxVnizHBlPRRF/EAc0KHlM8uEb8hBg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-22T08:37:07.030122Z","bundle_sha256":"ba5f24267e78e7bcb786a66c21b475116ded3a1e1f0a80a405a3109f5f16ae54"}}