{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:IYT3OGXGEAZAJWQQ46LZ63IYM5","short_pith_number":"pith:IYT3OGXG","canonical_record":{"source":{"id":"2502.04040","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-06T13:01:44Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"f78af78dda5b63479c48e9118054f88f906fc91a6ff452c403625138aba0ab63","abstract_canon_sha256":"8e1db630a0ae4d619b2a900634b065df2ece4b2ef82c6009131846b1673afc96"},"schema_version":"1.0"},"canonical_sha256":"4627b71ae6203204da10e7979f6d1867595d19bb74e2aa16a9f0b27741a2c658","source":{"kind":"arxiv","id":"2502.04040","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.04040","created_at":"2026-07-05T11:12:25Z"},{"alias_kind":"arxiv_version","alias_value":"2502.04040v2","created_at":"2026-07-05T11:12:25Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.04040","created_at":"2026-07-05T11:12:25Z"},{"alias_kind":"pith_short_12","alias_value":"IYT3OGXGEAZA","created_at":"2026-07-05T11:12:25Z"},{"alias_kind":"pith_short_16","alias_value":"IYT3OGXGEAZAJWQQ","created_at":"2026-07-05T11:12:25Z"},{"alias_kind":"pith_short_8","alias_value":"IYT3OGXG","created_at":"2026-07-05T11:12:25Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:IYT3OGXGEAZAJWQQ46LZ63IYM5","target":"record","payload":{"canonical_record":{"source":{"id":"2502.04040","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-06T13:01:44Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"f78af78dda5b63479c48e9118054f88f906fc91a6ff452c403625138aba0ab63","abstract_canon_sha256":"8e1db630a0ae4d619b2a900634b065df2ece4b2ef82c6009131846b1673afc96"},"schema_version":"1.0"},"canonical_sha256":"4627b71ae6203204da10e7979f6d1867595d19bb74e2aa16a9f0b27741a2c658","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:12:25.593014Z","signature_b64":"SIkExC9lUWrZM6RHHs+xpOGYWrSEnWfgfNg3nABc72aCIJTODBTvXXPjiEVhecGFoL3XLIw8D/Myod4TzBiuBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4627b71ae6203204da10e7979f6d1867595d19bb74e2aa16a9f0b27741a2c658","last_reissued_at":"2026-07-05T11:12:25.592524Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:12:25.592524Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2502.04040","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":"2026-07-05T11:12:25Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"UiQ3TX8fmpk20TqNaZgHbWkxq5sGWuNDHuhnMzTPQhVh0qtR85lkJ3TJMelfuFEsrMIbGX4b428bPO/ydSO7DQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-19T01:08:47.437586Z"},"content_sha256":"30cfafe5694088b30d8b5816627ad46af228629cfea7e22c81354c16a57bb453","schema_version":"1.0","event_id":"sha256:30cfafe5694088b30d8b5816627ad46af228629cfea7e22c81354c16a57bb453"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:IYT3OGXGEAZAJWQQ46LZ63IYM5","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Safety Reasoning with Guidelines","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Dacheng Tao, Haoyu Wang, Li Shen, Minhao Cheng, Xueqian Wang, Zeyu Qin","submitted_at":"2025-02-06T13:01:44Z","abstract_excerpt":"Training safe LLMs remains a critical challenge. The most widely used method, Refusal Training (RT), struggles to generalize against various Out-of-Distribution (OOD) jailbreaking attacks. Although various advanced methods have been proposed to address this issue, we instead question whether OOD attacks inherently surpass the capability of vanilla RT. Evaluations using Best-of-N (BoN) reveal significant safety improvements as N increases, indicating models possess adequate latent safety knowledge but RT fails to consistently elicit it under OOD scenarios. Further domain adaptation analysis rev"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.04040","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/2502.04040/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":"2026-07-05T11:12:25Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ryz8s+aah2SxiSYvzsglG0rkCz4n4j4Mvp4B+oieV4fhNKZIX4Mws1Jz+WzUPpAutQSi2p9Bu6IZUx7z141QAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-19T01:08:47.438158Z"},"content_sha256":"c9000a8963313e45b5e3db3b60139d9a3e0628388171aa50d85b0cc411eeeb74","schema_version":"1.0","event_id":"sha256:c9000a8963313e45b5e3db3b60139d9a3e0628388171aa50d85b0cc411eeeb74"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/IYT3OGXGEAZAJWQQ46LZ63IYM5/bundle.json","state_url":"https://pith.science/pith/IYT3OGXGEAZAJWQQ46LZ63IYM5/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/IYT3OGXGEAZAJWQQ46LZ63IYM5/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":"2026-08-19T01:08:47Z","links":{"resolver":"https://pith.science/pith/IYT3OGXGEAZAJWQQ46LZ63IYM5","bundle":"https://pith.science/pith/IYT3OGXGEAZAJWQQ46LZ63IYM5/bundle.json","state":"https://pith.science/pith/IYT3OGXGEAZAJWQQ46LZ63IYM5/state.json","well_known_bundle":"https://pith.science/.well-known/pith/IYT3OGXGEAZAJWQQ46LZ63IYM5/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:IYT3OGXGEAZAJWQQ46LZ63IYM5","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":"8e1db630a0ae4d619b2a900634b065df2ece4b2ef82c6009131846b1673afc96","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-06T13:01:44Z","title_canon_sha256":"f78af78dda5b63479c48e9118054f88f906fc91a6ff452c403625138aba0ab63"},"schema_version":"1.0","source":{"id":"2502.04040","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.04040","created_at":"2026-07-05T11:12:25Z"},{"alias_kind":"arxiv_version","alias_value":"2502.04040v2","created_at":"2026-07-05T11:12:25Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.04040","created_at":"2026-07-05T11:12:25Z"},{"alias_kind":"pith_short_12","alias_value":"IYT3OGXGEAZA","created_at":"2026-07-05T11:12:25Z"},{"alias_kind":"pith_short_16","alias_value":"IYT3OGXGEAZAJWQQ","created_at":"2026-07-05T11:12:25Z"},{"alias_kind":"pith_short_8","alias_value":"IYT3OGXG","created_at":"2026-07-05T11:12:25Z"}],"graph_snapshots":[{"event_id":"sha256:c9000a8963313e45b5e3db3b60139d9a3e0628388171aa50d85b0cc411eeeb74","target":"graph","created_at":"2026-07-05T11:12:25Z","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/2502.04040/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Training safe LLMs remains a critical challenge. The most widely used method, Refusal Training (RT), struggles to generalize against various Out-of-Distribution (OOD) jailbreaking attacks. Although various advanced methods have been proposed to address this issue, we instead question whether OOD attacks inherently surpass the capability of vanilla RT. Evaluations using Best-of-N (BoN) reveal significant safety improvements as N increases, indicating models possess adequate latent safety knowledge but RT fails to consistently elicit it under OOD scenarios. Further domain adaptation analysis rev","authors_text":"Dacheng Tao, Haoyu Wang, Li Shen, Minhao Cheng, Xueqian Wang, Zeyu Qin","cross_cats":["cs.AI","cs.CL"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-06T13:01:44Z","title":"Safety Reasoning with Guidelines"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.04040","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":"sha256:30cfafe5694088b30d8b5816627ad46af228629cfea7e22c81354c16a57bb453","target":"record","created_at":"2026-07-05T11:12:25Z","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":"8e1db630a0ae4d619b2a900634b065df2ece4b2ef82c6009131846b1673afc96","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-06T13:01:44Z","title_canon_sha256":"f78af78dda5b63479c48e9118054f88f906fc91a6ff452c403625138aba0ab63"},"schema_version":"1.0","source":{"id":"2502.04040","kind":"arxiv","version":2}},"canonical_sha256":"4627b71ae6203204da10e7979f6d1867595d19bb74e2aa16a9f0b27741a2c658","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"4627b71ae6203204da10e7979f6d1867595d19bb74e2aa16a9f0b27741a2c658","first_computed_at":"2026-07-05T11:12:25.592524Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:12:25.592524Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"SIkExC9lUWrZM6RHHs+xpOGYWrSEnWfgfNg3nABc72aCIJTODBTvXXPjiEVhecGFoL3XLIw8D/Myod4TzBiuBw==","signature_status":"signed_v1","signed_at":"2026-07-05T11:12:25.593014Z","signed_message":"canonical_sha256_bytes"},"source_id":"2502.04040","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:30cfafe5694088b30d8b5816627ad46af228629cfea7e22c81354c16a57bb453","sha256:c9000a8963313e45b5e3db3b60139d9a3e0628388171aa50d85b0cc411eeeb74"],"state_sha256":"62e09fc578ba8fdd8f643e5315d26d980d5a45c4bec4afcad3a8cf2bf9f65da2"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"8+w0stqKgwEEawnXM7CzA2poYLTgRj4zqYnNGBls8b6GZ0EfDCJ20ZoaMX1WvIQDjGOROnCH/MXmTeRrmAGXAw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-19T01:08:47.443043Z","bundle_sha256":"07391a4a4cf23336cfdf6b42511586e0bd6ab7c5a8cfbb33c1d69dbc1d14cc0e"}}