{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:WNLHMCPU3H4JIZ6DE27B57DOUE","short_pith_number":"pith:WNLHMCPU","schema_version":"1.0","canonical_sha256":"b3567609f4d9f89467c326be1efc6ea13a8cd516d9c9ecef882362630ff25532","source":{"kind":"arxiv","id":"2606.30899","version":1},"attestation_state":"computed","paper":{"title":"Curvature-Guided Module Localization for Low-Rank Detoxification of Backdoored Large Language Models","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Andrew Arash Mahyari, Arash Raftari, Mehrdad Mahdavi, Nathan Blackthorn","submitted_at":"2026-06-29T20:40:21Z","abstract_excerpt":"Backdoor attacks pose a serious threat to large language models (LLMs) by causing otherwise benign systems to produce attacker-specified malicious behavior when a hidden trigger is present. In this work, we study post hoc detoxification of backdoored LLMs in a practical setting where the defender has access to the poisoned model but does not wish to retrain the full network from scratch. We propose a mechanistically guided weight-space repair framework that first localizes modules involved in propagating trigger-induced behavior using activation patching and Fisher/K-FAC curvature analysis, an"},"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":"2606.30899","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CR","submitted_at":"2026-06-29T20:40:21Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"b88f8edd953240022685755778c3610b15f7179f3281a43eeb90442cfbb02d97","abstract_canon_sha256":"74ee4a4804ef4e80206410d1dc786e83d4788f32a4962c8808e09c840da04a77"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-01T00:17:20.872902Z","signature_b64":"9aVKLPWhPBq0yXHIotkqnAQev+1/1mdKuaOfanOPglTvR3RzcUw9as3ALtPNKvGQ7uF4+DNxQgCrIqsRkTMABQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b3567609f4d9f89467c326be1efc6ea13a8cd516d9c9ecef882362630ff25532","last_reissued_at":"2026-07-01T00:17:20.872445Z","signature_status":"signed_v1","first_computed_at":"2026-07-01T00:17:20.872445Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Curvature-Guided Module Localization for Low-Rank Detoxification of Backdoored Large Language Models","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Andrew Arash Mahyari, Arash Raftari, Mehrdad Mahdavi, Nathan Blackthorn","submitted_at":"2026-06-29T20:40:21Z","abstract_excerpt":"Backdoor attacks pose a serious threat to large language models (LLMs) by causing otherwise benign systems to produce attacker-specified malicious behavior when a hidden trigger is present. In this work, we study post hoc detoxification of backdoored LLMs in a practical setting where the defender has access to the poisoned model but does not wish to retrain the full network from scratch. We propose a mechanistically guided weight-space repair framework that first localizes modules involved in propagating trigger-induced behavior using activation patching and Fisher/K-FAC curvature analysis, an"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.30899","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/2606.30899/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":"2606.30899","created_at":"2026-07-01T00:17:20.872508+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.30899v1","created_at":"2026-07-01T00:17:20.872508+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.30899","created_at":"2026-07-01T00:17:20.872508+00:00"},{"alias_kind":"pith_short_12","alias_value":"WNLHMCPU3H4J","created_at":"2026-07-01T00:17:20.872508+00:00"},{"alias_kind":"pith_short_16","alias_value":"WNLHMCPU3H4JIZ6D","created_at":"2026-07-01T00:17:20.872508+00:00"},{"alias_kind":"pith_short_8","alias_value":"WNLHMCPU","created_at":"2026-07-01T00:17:20.872508+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/WNLHMCPU3H4JIZ6DE27B57DOUE","json":"https://pith.science/pith/WNLHMCPU3H4JIZ6DE27B57DOUE.json","graph_json":"https://pith.science/api/pith-number/WNLHMCPU3H4JIZ6DE27B57DOUE/graph.json","events_json":"https://pith.science/api/pith-number/WNLHMCPU3H4JIZ6DE27B57DOUE/events.json","paper":"https://pith.science/paper/WNLHMCPU"},"agent_actions":{"view_html":"https://pith.science/pith/WNLHMCPU3H4JIZ6DE27B57DOUE","download_json":"https://pith.science/pith/WNLHMCPU3H4JIZ6DE27B57DOUE.json","view_paper":"https://pith.science/paper/WNLHMCPU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.30899&json=true","fetch_graph":"https://pith.science/api/pith-number/WNLHMCPU3H4JIZ6DE27B57DOUE/graph.json","fetch_events":"https://pith.science/api/pith-number/WNLHMCPU3H4JIZ6DE27B57DOUE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WNLHMCPU3H4JIZ6DE27B57DOUE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WNLHMCPU3H4JIZ6DE27B57DOUE/action/storage_attestation","attest_author":"https://pith.science/pith/WNLHMCPU3H4JIZ6DE27B57DOUE/action/author_attestation","sign_citation":"https://pith.science/pith/WNLHMCPU3H4JIZ6DE27B57DOUE/action/citation_signature","submit_replication":"https://pith.science/pith/WNLHMCPU3H4JIZ6DE27B57DOUE/action/replication_record"}},"created_at":"2026-07-01T00:17:20.872508+00:00","updated_at":"2026-07-01T00:17:20.872508+00:00"}