{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:UDVPJ3SOKTTWDP4PMEKS4L3MPB","short_pith_number":"pith:UDVPJ3SO","canonical_record":{"source":{"id":"2410.09760","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-10-13T07:36:45Z","cross_cats_sorted":[],"title_canon_sha256":"4ca1b23e57245cfac1f4b6bee86a2d52c41b2e50dc6410fabf132361a6753a1b","abstract_canon_sha256":"d97337f48dec872ea6f53b399645d2cab8030048d806b44bf3ee06b6192e8632"},"schema_version":"1.0"},"canonical_sha256":"a0eaf4ee4e54e761bf8f61152e2f6c7851b5de788213425f4c417b337911ec92","source":{"kind":"arxiv","id":"2410.09760","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2410.09760","created_at":"2026-07-05T10:07:45Z"},{"alias_kind":"arxiv_version","alias_value":"2410.09760v3","created_at":"2026-07-05T10:07:45Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.09760","created_at":"2026-07-05T10:07:45Z"},{"alias_kind":"pith_short_12","alias_value":"UDVPJ3SOKTTW","created_at":"2026-07-05T10:07:45Z"},{"alias_kind":"pith_short_16","alias_value":"UDVPJ3SOKTTWDP4P","created_at":"2026-07-05T10:07:45Z"},{"alias_kind":"pith_short_8","alias_value":"UDVPJ3SO","created_at":"2026-07-05T10:07:45Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:UDVPJ3SOKTTWDP4PMEKS4L3MPB","target":"record","payload":{"canonical_record":{"source":{"id":"2410.09760","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-10-13T07:36:45Z","cross_cats_sorted":[],"title_canon_sha256":"4ca1b23e57245cfac1f4b6bee86a2d52c41b2e50dc6410fabf132361a6753a1b","abstract_canon_sha256":"d97337f48dec872ea6f53b399645d2cab8030048d806b44bf3ee06b6192e8632"},"schema_version":"1.0"},"canonical_sha256":"a0eaf4ee4e54e761bf8f61152e2f6c7851b5de788213425f4c417b337911ec92","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:07:45.388453Z","signature_b64":"aGV/PXLpNmvfziLLLphQHla+Hh6WPHemGnP8JBeUA6h9Z3jRR3CiycUK9SvcZyclK25QpuLxEm15+7WHIfyjDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a0eaf4ee4e54e761bf8f61152e2f6c7851b5de788213425f4c417b337911ec92","last_reissued_at":"2026-07-05T10:07:45.388018Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:07:45.388018Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2410.09760","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":"2026-07-05T10:07:45Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"t91PLAk71Iq+JXH1018wYfSV3rIDHlOk2XYICRqeYf28tjD1Q6G2K6nNQu//+0fqoiagCTWkE02phN5ODv6nAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T23:32:44.183166Z"},"content_sha256":"d9e8848148f7d9ea44ad0fe0054a3f8a3134f734c6a00367f442fc56635118c4","schema_version":"1.0","event_id":"sha256:d9e8848148f7d9ea44ad0fe0054a3f8a3134f734c6a00367f442fc56635118c4"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:UDVPJ3SOKTTWDP4PMEKS4L3MPB","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Targeted Vaccine: Safety Alignment for Large Language Models against Harmful Fine-Tuning via Layer-wise Perturbation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Guozhi Liu, Li Shen, Qi Mu, Ruichao Mo, Tiansheng Huang, Weiwei Lin","submitted_at":"2024-10-13T07:36:45Z","abstract_excerpt":"Harmful fine-tuning attack poses a serious threat to the online fine-tuning service. Vaccine, a recent alignment-stage defense, applies uniform perturbation to all layers of embedding to make the model robust to the simulated embedding drift. However, applying layer-wise uniform perturbation may lead to excess perturbations for some particular safety-irrelevant layers, resulting in defense performance degradation and unnecessary memory consumption. To address this limitation, we propose Targeted Vaccine (T-Vaccine), a memory-efficient safety alignment method that applies perturbation to only s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.09760","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/2410.09760/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":"2026-07-05T10:07:45Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"iiG8Cc8cpjqls9A7Pt//Ab2rkalahiHX83nAkWCV3UHVhcZsNB4igAvP+wydiMsfV6N6q8fcqg+Imk2gcNFRBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T23:32:44.184757Z"},"content_sha256":"836f42f856799141920fb6b8727d49edd674debe25ae536db5a78b19dfdffad6","schema_version":"1.0","event_id":"sha256:836f42f856799141920fb6b8727d49edd674debe25ae536db5a78b19dfdffad6"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/UDVPJ3SOKTTWDP4PMEKS4L3MPB/bundle.json","state_url":"https://pith.science/pith/UDVPJ3SOKTTWDP4PMEKS4L3MPB/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/UDVPJ3SOKTTWDP4PMEKS4L3MPB/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":"2026-08-04T23:32:44Z","links":{"resolver":"https://pith.science/pith/UDVPJ3SOKTTWDP4PMEKS4L3MPB","bundle":"https://pith.science/pith/UDVPJ3SOKTTWDP4PMEKS4L3MPB/bundle.json","state":"https://pith.science/pith/UDVPJ3SOKTTWDP4PMEKS4L3MPB/state.json","well_known_bundle":"https://pith.science/.well-known/pith/UDVPJ3SOKTTWDP4PMEKS4L3MPB/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:UDVPJ3SOKTTWDP4PMEKS4L3MPB","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":"d97337f48dec872ea6f53b399645d2cab8030048d806b44bf3ee06b6192e8632","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-10-13T07:36:45Z","title_canon_sha256":"4ca1b23e57245cfac1f4b6bee86a2d52c41b2e50dc6410fabf132361a6753a1b"},"schema_version":"1.0","source":{"id":"2410.09760","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2410.09760","created_at":"2026-07-05T10:07:45Z"},{"alias_kind":"arxiv_version","alias_value":"2410.09760v3","created_at":"2026-07-05T10:07:45Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.09760","created_at":"2026-07-05T10:07:45Z"},{"alias_kind":"pith_short_12","alias_value":"UDVPJ3SOKTTW","created_at":"2026-07-05T10:07:45Z"},{"alias_kind":"pith_short_16","alias_value":"UDVPJ3SOKTTWDP4P","created_at":"2026-07-05T10:07:45Z"},{"alias_kind":"pith_short_8","alias_value":"UDVPJ3SO","created_at":"2026-07-05T10:07:45Z"}],"graph_snapshots":[{"event_id":"sha256:836f42f856799141920fb6b8727d49edd674debe25ae536db5a78b19dfdffad6","target":"graph","created_at":"2026-07-05T10:07:45Z","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/2410.09760/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Harmful fine-tuning attack poses a serious threat to the online fine-tuning service. Vaccine, a recent alignment-stage defense, applies uniform perturbation to all layers of embedding to make the model robust to the simulated embedding drift. However, applying layer-wise uniform perturbation may lead to excess perturbations for some particular safety-irrelevant layers, resulting in defense performance degradation and unnecessary memory consumption. To address this limitation, we propose Targeted Vaccine (T-Vaccine), a memory-efficient safety alignment method that applies perturbation to only s","authors_text":"Guozhi Liu, Li Shen, Qi Mu, Ruichao Mo, Tiansheng Huang, Weiwei Lin","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-10-13T07:36:45Z","title":"Targeted Vaccine: Safety Alignment for Large Language Models against Harmful Fine-Tuning via Layer-wise Perturbation"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.09760","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:d9e8848148f7d9ea44ad0fe0054a3f8a3134f734c6a00367f442fc56635118c4","target":"record","created_at":"2026-07-05T10:07:45Z","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":"d97337f48dec872ea6f53b399645d2cab8030048d806b44bf3ee06b6192e8632","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-10-13T07:36:45Z","title_canon_sha256":"4ca1b23e57245cfac1f4b6bee86a2d52c41b2e50dc6410fabf132361a6753a1b"},"schema_version":"1.0","source":{"id":"2410.09760","kind":"arxiv","version":3}},"canonical_sha256":"a0eaf4ee4e54e761bf8f61152e2f6c7851b5de788213425f4c417b337911ec92","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a0eaf4ee4e54e761bf8f61152e2f6c7851b5de788213425f4c417b337911ec92","first_computed_at":"2026-07-05T10:07:45.388018Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:07:45.388018Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"aGV/PXLpNmvfziLLLphQHla+Hh6WPHemGnP8JBeUA6h9Z3jRR3CiycUK9SvcZyclK25QpuLxEm15+7WHIfyjDQ==","signature_status":"signed_v1","signed_at":"2026-07-05T10:07:45.388453Z","signed_message":"canonical_sha256_bytes"},"source_id":"2410.09760","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:d9e8848148f7d9ea44ad0fe0054a3f8a3134f734c6a00367f442fc56635118c4","sha256:836f42f856799141920fb6b8727d49edd674debe25ae536db5a78b19dfdffad6"],"state_sha256":"de8b5fcb38992c086f04a204ba1445b412ef4b00c44d4fcbf04106da99c2077e"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ml5H8KgTe0hF0dPZ7k+RwpSZyzG6kOUvsMwy5nPaRlZl1WCzZ0uyCT1n7G1G2NGelIc/C1az3RfB7LlwbhOYBg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-04T23:32:44.192741Z","bundle_sha256":"5cbec04354aa197ca7c8395456f95a7605a356da93240c61951f36392d112c23"}}