{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ZX4ZA4QLXX5D2FDK2HITLLJRFY","short_pith_number":"pith:ZX4ZA4QL","schema_version":"1.0","canonical_sha256":"cdf990720bbdfa3d146ad1d135ad312e2d9512a86ac67367d5da7957d4f3d46e","source":{"kind":"arxiv","id":"2405.16229","version":1},"attestation_state":"computed","paper":{"title":"No Two Devils Alike: Unveiling Distinct Mechanisms of Fine-tuning Attacks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.CL","authors_text":"Chak Tou Leong, Hanlin Wang, Jian Wang, Kaishuai Xu, Wenjie Li, Yi Cheng","submitted_at":"2024-05-25T13:38:40Z","abstract_excerpt":"The existing safety alignment of Large Language Models (LLMs) is found fragile and could be easily attacked through different strategies, such as through fine-tuning on a few harmful examples or manipulating the prefix of the generation results. However, the attack mechanisms of these strategies are still underexplored. In this paper, we ask the following question: \\textit{while these approaches can all significantly compromise safety, do their attack mechanisms exhibit strong similarities?} To answer this question, we break down the safeguarding process of an LLM when encountered with harmful"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2405.16229","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-25T13:38:40Z","cross_cats_sorted":["cs.CR"],"title_canon_sha256":"89ea185486146241766b8ab0f2809b4b1f6a3484dbd548b48d56808d8faace15","abstract_canon_sha256":"79230b68b3034da342dff89d54e2b51a9d0d8e56927e5cb0730b4d3f404860a1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:23:27.511618Z","signature_b64":"fDVn4v6hd793dxTMrwfQa9ybFEOWToYnREquh4/VDAGkvsybvVe5gp4Kn8Ra5yzhDnjg0Df9HoEl7DFR60D3CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cdf990720bbdfa3d146ad1d135ad312e2d9512a86ac67367d5da7957d4f3d46e","last_reissued_at":"2026-07-05T08:23:27.511043Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:23:27.511043Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"No Two Devils Alike: Unveiling Distinct Mechanisms of Fine-tuning Attacks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.CL","authors_text":"Chak Tou Leong, Hanlin Wang, Jian Wang, Kaishuai Xu, Wenjie Li, Yi Cheng","submitted_at":"2024-05-25T13:38:40Z","abstract_excerpt":"The existing safety alignment of Large Language Models (LLMs) is found fragile and could be easily attacked through different strategies, such as through fine-tuning on a few harmful examples or manipulating the prefix of the generation results. However, the attack mechanisms of these strategies are still underexplored. In this paper, we ask the following question: \\textit{while these approaches can all significantly compromise safety, do their attack mechanisms exhibit strong similarities?} To answer this question, we break down the safeguarding process of an LLM when encountered with harmful"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.16229","kind":"arxiv","version":1},"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/2405.16229/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2405.16229","created_at":"2026-07-05T08:23:27.511108+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.16229v1","created_at":"2026-07-05T08:23:27.511108+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.16229","created_at":"2026-07-05T08:23:27.511108+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZX4ZA4QLXX5D","created_at":"2026-07-05T08:23:27.511108+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZX4ZA4QLXX5D2FDK","created_at":"2026-07-05T08:23:27.511108+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZX4ZA4QL","created_at":"2026-07-05T08:23:27.511108+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2409.18169","citing_title":"Harmful Fine-tuning Attacks and Defenses for Large Language Models: A Survey","ref_index":84,"is_internal_anchor":false},{"citing_arxiv_id":"2404.01318","citing_title":"JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18510","citing_title":"Different Paths to Harmful Compliance: Behavioral Side Effects and Mechanistic Divergence Across LLM Jailbreaks","ref_index":15,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZX4ZA4QLXX5D2FDK2HITLLJRFY","json":"https://pith.science/pith/ZX4ZA4QLXX5D2FDK2HITLLJRFY.json","graph_json":"https://pith.science/api/pith-number/ZX4ZA4QLXX5D2FDK2HITLLJRFY/graph.json","events_json":"https://pith.science/api/pith-number/ZX4ZA4QLXX5D2FDK2HITLLJRFY/events.json","paper":"https://pith.science/paper/ZX4ZA4QL"},"agent_actions":{"view_html":"https://pith.science/pith/ZX4ZA4QLXX5D2FDK2HITLLJRFY","download_json":"https://pith.science/pith/ZX4ZA4QLXX5D2FDK2HITLLJRFY.json","view_paper":"https://pith.science/paper/ZX4ZA4QL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.16229&json=true","fetch_graph":"https://pith.science/api/pith-number/ZX4ZA4QLXX5D2FDK2HITLLJRFY/graph.json","fetch_events":"https://pith.science/api/pith-number/ZX4ZA4QLXX5D2FDK2HITLLJRFY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZX4ZA4QLXX5D2FDK2HITLLJRFY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZX4ZA4QLXX5D2FDK2HITLLJRFY/action/storage_attestation","attest_author":"https://pith.science/pith/ZX4ZA4QLXX5D2FDK2HITLLJRFY/action/author_attestation","sign_citation":"https://pith.science/pith/ZX4ZA4QLXX5D2FDK2HITLLJRFY/action/citation_signature","submit_replication":"https://pith.science/pith/ZX4ZA4QLXX5D2FDK2HITLLJRFY/action/replication_record"}},"created_at":"2026-07-05T08:23:27.511108+00:00","updated_at":"2026-07-05T08:23:27.511108+00:00"}