{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:VS74OPA6Z6M2SXOTLRY47AGCSS","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":"8cafc6f2c643ef58b7d5ca99915224c082e5a4ec96f8e0ed4c4dc412c313d93e","cross_cats_sorted":["cs.CR"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-08-18T21:45:03Z","title_canon_sha256":"cc677f136fd792f9f37796162259c7ed5ea0044facd8b5a835f8bc934b2d79d6"},"schema_version":"1.0","source":{"id":"2408.09600","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2408.09600","created_at":"2026-07-05T12:05:09Z"},{"alias_kind":"arxiv_version","alias_value":"2408.09600v3","created_at":"2026-07-05T12:05:09Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.09600","created_at":"2026-07-05T12:05:09Z"},{"alias_kind":"pith_short_12","alias_value":"VS74OPA6Z6M2","created_at":"2026-07-05T12:05:09Z"},{"alias_kind":"pith_short_16","alias_value":"VS74OPA6Z6M2SXOT","created_at":"2026-07-05T12:05:09Z"},{"alias_kind":"pith_short_8","alias_value":"VS74OPA6","created_at":"2026-07-05T12:05:09Z"}],"graph_snapshots":[{"event_id":"sha256:0e8a6ffed3d397c85822dcf3406abfa7098ee18a6608a9418a2a26f43f4e32f2","target":"graph","created_at":"2026-07-05T12:05:09Z","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/2408.09600/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Safety aligned Large Language Models (LLMs) are vulnerable to harmful fine-tuning attacks -- a few harmful data mixed in the fine-tuning dataset can break the LLMs's safety alignment. While several defenses have been proposed, our evaluation shows that existing defenses fail \\textit{when some specific training hyper-parameters are chosen} -- a large learning rate or a large number of training epochs in the fine-tuning stage can easily invalidate the defense. To this end, we propose Antidote, a post-fine-tuning stage solution, which remains \\textbf{\\textit{agnostic to the training hyper-paramet","authors_text":"Gautam Bhattacharya, Josh Kimball, Ling Liu, Pratik Joshi, Tiansheng Huang","cross_cats":["cs.CR"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-08-18T21:45:03Z","title":"Antidote: Post-fine-tuning Safety Alignment for Large Language Models against Harmful Fine-tuning"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2408.09600","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":"sha256:302ce7404779525ed9aa52437b5d89598a76ffdb8a10eea188ebc992b0cd55c8","target":"record","created_at":"2026-07-05T12:05:09Z","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":"8cafc6f2c643ef58b7d5ca99915224c082e5a4ec96f8e0ed4c4dc412c313d93e","cross_cats_sorted":["cs.CR"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-08-18T21:45:03Z","title_canon_sha256":"cc677f136fd792f9f37796162259c7ed5ea0044facd8b5a835f8bc934b2d79d6"},"schema_version":"1.0","source":{"id":"2408.09600","kind":"arxiv","version":3}},"canonical_sha256":"acbfc73c1ecf99a95dd35c71cf80c29485f82367361cb88d71be3ae2a04f7f1e","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"acbfc73c1ecf99a95dd35c71cf80c29485f82367361cb88d71be3ae2a04f7f1e","first_computed_at":"2026-07-05T12:05:09.791778Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T12:05:09.791778Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"6vgFt8tdcJl4v+A+T8EiImgtN91CGnoOVtYzYw2suLB4YsF/qiJyvPjIpDo/44zPEreOpaAn5TyUtwEZ1x6LDg==","signature_status":"signed_v1","signed_at":"2026-07-05T12:05:09.792311Z","signed_message":"canonical_sha256_bytes"},"source_id":"2408.09600","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:302ce7404779525ed9aa52437b5d89598a76ffdb8a10eea188ebc992b0cd55c8","sha256:0e8a6ffed3d397c85822dcf3406abfa7098ee18a6608a9418a2a26f43f4e32f2"],"state_sha256":"d69aca2a01ca62d7eb02baf00ec7a1d6fb6da01723d14b11e59db412d6fc6d61"}