{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:2H5UPSO3GQ6KOFSD66PM4RFZXX","short_pith_number":"pith:2H5UPSO3","schema_version":"1.0","canonical_sha256":"d1fb47c9db343ca71643f79ece44b9bdc84059271d3b6105ad1407217c36b9de","source":{"kind":"arxiv","id":"2412.12497","version":1},"attestation_state":"computed","paper":{"title":"NLSR: Neuron-Level Safety Realignment of Large Language Models Against Harmful Fine-Tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Gerard de Melo, Liang He, Linlin Wang, Shunfan Zheng, Xiaoling Wang, Xin Yi","submitted_at":"2024-12-17T02:59:04Z","abstract_excerpt":"The emergence of finetuning-as-a-service has revealed a new vulnerability in large language models (LLMs). A mere handful of malicious data uploaded by users can subtly manipulate the finetuning process, resulting in an alignment-broken model. Existing methods to counteract fine-tuning attacks typically require substantial computational resources. Even with parameter-efficient techniques like LoRA, gradient updates remain essential. To address these challenges, we propose \\textbf{N}euron-\\textbf{L}evel \\textbf{S}afety \\textbf{R}ealignment (\\textbf{NLSR}), a training-free framework that restore"},"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":"2412.12497","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-12-17T02:59:04Z","cross_cats_sorted":[],"title_canon_sha256":"9fe8e07b08fb15c7d43f01c2c2ffff2cd6c9894f359b1fd4e7e0d06728fca9c5","abstract_canon_sha256":"8145211db01004c69a6fbe9765b24a475895daeca87b4b6a6ab21271faf03e24"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:50:15.883397Z","signature_b64":"/bC8pE5+O1j98DMUJqvVxc042j7x2Ls7oLp4ic3sokeJ5TRU2P27ekGHxnQKb5GoHaom8IYrCi8EAsUa3LsuBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d1fb47c9db343ca71643f79ece44b9bdc84059271d3b6105ad1407217c36b9de","last_reissued_at":"2026-07-05T09:50:15.882855Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:50:15.882855Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"NLSR: Neuron-Level Safety Realignment of Large Language Models Against Harmful Fine-Tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Gerard de Melo, Liang He, Linlin Wang, Shunfan Zheng, Xiaoling Wang, Xin Yi","submitted_at":"2024-12-17T02:59:04Z","abstract_excerpt":"The emergence of finetuning-as-a-service has revealed a new vulnerability in large language models (LLMs). A mere handful of malicious data uploaded by users can subtly manipulate the finetuning process, resulting in an alignment-broken model. Existing methods to counteract fine-tuning attacks typically require substantial computational resources. Even with parameter-efficient techniques like LoRA, gradient updates remain essential. To address these challenges, we propose \\textbf{N}euron-\\textbf{L}evel \\textbf{S}afety \\textbf{R}ealignment (\\textbf{NLSR}), a training-free framework that restore"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.12497","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/2412.12497/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":"2412.12497","created_at":"2026-07-05T09:50:15.882915+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.12497v1","created_at":"2026-07-05T09:50:15.882915+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.12497","created_at":"2026-07-05T09:50:15.882915+00:00"},{"alias_kind":"pith_short_12","alias_value":"2H5UPSO3GQ6K","created_at":"2026-07-05T09:50:15.882915+00:00"},{"alias_kind":"pith_short_16","alias_value":"2H5UPSO3GQ6KOFSD","created_at":"2026-07-05T09:50:15.882915+00:00"},{"alias_kind":"pith_short_8","alias_value":"2H5UPSO3","created_at":"2026-07-05T09:50:15.882915+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2H5UPSO3GQ6KOFSD66PM4RFZXX","json":"https://pith.science/pith/2H5UPSO3GQ6KOFSD66PM4RFZXX.json","graph_json":"https://pith.science/api/pith-number/2H5UPSO3GQ6KOFSD66PM4RFZXX/graph.json","events_json":"https://pith.science/api/pith-number/2H5UPSO3GQ6KOFSD66PM4RFZXX/events.json","paper":"https://pith.science/paper/2H5UPSO3"},"agent_actions":{"view_html":"https://pith.science/pith/2H5UPSO3GQ6KOFSD66PM4RFZXX","download_json":"https://pith.science/pith/2H5UPSO3GQ6KOFSD66PM4RFZXX.json","view_paper":"https://pith.science/paper/2H5UPSO3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.12497&json=true","fetch_graph":"https://pith.science/api/pith-number/2H5UPSO3GQ6KOFSD66PM4RFZXX/graph.json","fetch_events":"https://pith.science/api/pith-number/2H5UPSO3GQ6KOFSD66PM4RFZXX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2H5UPSO3GQ6KOFSD66PM4RFZXX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2H5UPSO3GQ6KOFSD66PM4RFZXX/action/storage_attestation","attest_author":"https://pith.science/pith/2H5UPSO3GQ6KOFSD66PM4RFZXX/action/author_attestation","sign_citation":"https://pith.science/pith/2H5UPSO3GQ6KOFSD66PM4RFZXX/action/citation_signature","submit_replication":"https://pith.science/pith/2H5UPSO3GQ6KOFSD66PM4RFZXX/action/replication_record"}},"created_at":"2026-07-05T09:50:15.882915+00:00","updated_at":"2026-07-05T09:50:15.882915+00:00"}