{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:ODALYDF2OTUDXWHQZMHNZY4FPO","short_pith_number":"pith:ODALYDF2","canonical_record":{"source":{"id":"2402.05162","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-07T18:34:38Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"3344489afdccf9f16e936a88ebd47396f69ae75967ee7a9a707dab0cb3e721b4","abstract_canon_sha256":"54acf59667ff6ccfac9816ddab5f5c5d91785bff227fa2d91c6e6d70cf494667"},"schema_version":"1.0"},"canonical_sha256":"70c0bc0cba74e83bd8f0cb0edce3857ba0f99944f5681faf384d7bf7e7411d9c","source":{"kind":"arxiv","id":"2402.05162","version":4},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2402.05162","created_at":"2026-07-05T09:25:39Z"},{"alias_kind":"arxiv_version","alias_value":"2402.05162v4","created_at":"2026-07-05T09:25:39Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.05162","created_at":"2026-07-05T09:25:39Z"},{"alias_kind":"pith_short_12","alias_value":"ODALYDF2OTUD","created_at":"2026-07-05T09:25:39Z"},{"alias_kind":"pith_short_16","alias_value":"ODALYDF2OTUDXWHQ","created_at":"2026-07-05T09:25:39Z"},{"alias_kind":"pith_short_8","alias_value":"ODALYDF2","created_at":"2026-07-05T09:25:39Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:ODALYDF2OTUDXWHQZMHNZY4FPO","target":"record","payload":{"canonical_record":{"source":{"id":"2402.05162","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-07T18:34:38Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"3344489afdccf9f16e936a88ebd47396f69ae75967ee7a9a707dab0cb3e721b4","abstract_canon_sha256":"54acf59667ff6ccfac9816ddab5f5c5d91785bff227fa2d91c6e6d70cf494667"},"schema_version":"1.0"},"canonical_sha256":"70c0bc0cba74e83bd8f0cb0edce3857ba0f99944f5681faf384d7bf7e7411d9c","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:25:39.490603Z","signature_b64":"26B7orpxxgJZtiq0zHebfYh/xaN0yBUIuk3FuIXWsDo6Qln0qhHXUSUVZoXk1shu+/aBv/eJV7OdX34szhnABQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"70c0bc0cba74e83bd8f0cb0edce3857ba0f99944f5681faf384d7bf7e7411d9c","last_reissued_at":"2026-07-05T09:25:39.490118Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:25:39.490118Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2402.05162","source_version":4,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T09:25:39Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"djWuc+6dgkgW2Aa+AGl+j1Z0LHRzsnb3UOi4xg4zcG+jvuaU4p46fTBBLSLeIE1TeM0f95IAB70vJwCuF7HkDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-09T03:12:30.824036Z"},"content_sha256":"91e55cd19433177d8c5269103baa2279929add3f9e99b8c4c325f2005f6b475d","schema_version":"1.0","event_id":"sha256:91e55cd19433177d8c5269103baa2279929add3f9e99b8c4c325f2005f6b475d"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:ODALYDF2OTUDXWHQZMHNZY4FPO","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Assessing the Brittleness of Safety Alignment via Pruning and Low-Rank Modifications","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Boyi Wei, Kaixuan Huang, Mengdi Wang, Mengzhou Xia, Peter Henderson, Prateek Mittal, Tinghao Xie, Xiangyu Qi, Yangsibo Huang","submitted_at":"2024-02-07T18:34:38Z","abstract_excerpt":"Large language models (LLMs) show inherent brittleness in their safety mechanisms, as evidenced by their susceptibility to jailbreaking and even non-malicious fine-tuning. This study explores this brittleness of safety alignment by leveraging pruning and low-rank modifications. We develop methods to identify critical regions that are vital for safety guardrails, and that are disentangled from utility-relevant regions at both the neuron and rank levels. Surprisingly, the isolated regions we find are sparse, comprising about $3\\%$ at the parameter level and $2.5\\%$ at the rank level. Removing th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.05162","kind":"arxiv","version":4},"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/2402.05162/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":"2026-07-05T09:25:39Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Rry8PY4E8pX9Jaswlhbf4S/R7wIRPOUL5qwskD0t9zK6vX6jydPogxTmprmtLUehrcLmIKcPavgmKiNzi8wkBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-09T03:12:30.824419Z"},"content_sha256":"be724f5e90afe8f644e3b43098db7a75ae4298c8615c7392c38ae29c804cd3a5","schema_version":"1.0","event_id":"sha256:be724f5e90afe8f644e3b43098db7a75ae4298c8615c7392c38ae29c804cd3a5"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/ODALYDF2OTUDXWHQZMHNZY4FPO/bundle.json","state_url":"https://pith.science/pith/ODALYDF2OTUDXWHQZMHNZY4FPO/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/ODALYDF2OTUDXWHQZMHNZY4FPO/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":"2026-07-09T03:12:30Z","links":{"resolver":"https://pith.science/pith/ODALYDF2OTUDXWHQZMHNZY4FPO","bundle":"https://pith.science/pith/ODALYDF2OTUDXWHQZMHNZY4FPO/bundle.json","state":"https://pith.science/pith/ODALYDF2OTUDXWHQZMHNZY4FPO/state.json","well_known_bundle":"https://pith.science/.well-known/pith/ODALYDF2OTUDXWHQZMHNZY4FPO/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:ODALYDF2OTUDXWHQZMHNZY4FPO","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":"54acf59667ff6ccfac9816ddab5f5c5d91785bff227fa2d91c6e6d70cf494667","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-07T18:34:38Z","title_canon_sha256":"3344489afdccf9f16e936a88ebd47396f69ae75967ee7a9a707dab0cb3e721b4"},"schema_version":"1.0","source":{"id":"2402.05162","kind":"arxiv","version":4}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2402.05162","created_at":"2026-07-05T09:25:39Z"},{"alias_kind":"arxiv_version","alias_value":"2402.05162v4","created_at":"2026-07-05T09:25:39Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.05162","created_at":"2026-07-05T09:25:39Z"},{"alias_kind":"pith_short_12","alias_value":"ODALYDF2OTUD","created_at":"2026-07-05T09:25:39Z"},{"alias_kind":"pith_short_16","alias_value":"ODALYDF2OTUDXWHQ","created_at":"2026-07-05T09:25:39Z"},{"alias_kind":"pith_short_8","alias_value":"ODALYDF2","created_at":"2026-07-05T09:25:39Z"}],"graph_snapshots":[{"event_id":"sha256:be724f5e90afe8f644e3b43098db7a75ae4298c8615c7392c38ae29c804cd3a5","target":"graph","created_at":"2026-07-05T09:25:39Z","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/2402.05162/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Large language models (LLMs) show inherent brittleness in their safety mechanisms, as evidenced by their susceptibility to jailbreaking and even non-malicious fine-tuning. This study explores this brittleness of safety alignment by leveraging pruning and low-rank modifications. We develop methods to identify critical regions that are vital for safety guardrails, and that are disentangled from utility-relevant regions at both the neuron and rank levels. Surprisingly, the isolated regions we find are sparse, comprising about $3\\%$ at the parameter level and $2.5\\%$ at the rank level. Removing th","authors_text":"Boyi Wei, Kaixuan Huang, Mengdi Wang, Mengzhou Xia, Peter Henderson, Prateek Mittal, Tinghao Xie, Xiangyu Qi, Yangsibo Huang","cross_cats":["cs.AI","cs.CL"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-07T18:34:38Z","title":"Assessing the Brittleness of Safety Alignment via Pruning and Low-Rank Modifications"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.05162","kind":"arxiv","version":4},"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":"sha256:91e55cd19433177d8c5269103baa2279929add3f9e99b8c4c325f2005f6b475d","target":"record","created_at":"2026-07-05T09:25:39Z","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":"54acf59667ff6ccfac9816ddab5f5c5d91785bff227fa2d91c6e6d70cf494667","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-07T18:34:38Z","title_canon_sha256":"3344489afdccf9f16e936a88ebd47396f69ae75967ee7a9a707dab0cb3e721b4"},"schema_version":"1.0","source":{"id":"2402.05162","kind":"arxiv","version":4}},"canonical_sha256":"70c0bc0cba74e83bd8f0cb0edce3857ba0f99944f5681faf384d7bf7e7411d9c","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"70c0bc0cba74e83bd8f0cb0edce3857ba0f99944f5681faf384d7bf7e7411d9c","first_computed_at":"2026-07-05T09:25:39.490118Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T09:25:39.490118Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"26B7orpxxgJZtiq0zHebfYh/xaN0yBUIuk3FuIXWsDo6Qln0qhHXUSUVZoXk1shu+/aBv/eJV7OdX34szhnABQ==","signature_status":"signed_v1","signed_at":"2026-07-05T09:25:39.490603Z","signed_message":"canonical_sha256_bytes"},"source_id":"2402.05162","source_kind":"arxiv","source_version":4}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:91e55cd19433177d8c5269103baa2279929add3f9e99b8c4c325f2005f6b475d","sha256:be724f5e90afe8f644e3b43098db7a75ae4298c8615c7392c38ae29c804cd3a5"],"state_sha256":"6801f5ae7326896fb32d035492f218d64d0bb9dba7633972d49ab12d184c0017"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"IKjlGrCjoZQhP2t8jCNRUM5BcFdHn0Mq6QGr+sxgv0BZuUCRt83kQxZWpRqhPUknfri5ppFP0zDXpjMVxOfsBw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-09T03:12:30.826523Z","bundle_sha256":"c671509e48c2237dedec662dd220eaef7b571443bc8a74ccfcd972fd4329bc65"}}