{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:AY3HY2HY36GIQUYBYRYTGXUQFT","short_pith_number":"pith:AY3HY2HY","schema_version":"1.0","canonical_sha256":"06367c68f8df8c885301c471335e902cdb0346fd9a111517e3aac107a3a831d5","source":{"kind":"arxiv","id":"2606.03785","version":1},"attestation_state":"computed","paper":{"title":"Backdoor Unlearning Generalization: A Path Toward the Removal of Unknown Triggers in LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Djam\\'e Seddah, Lisa Bouger, Philippe Looubet Moundi, Th\\'eo Lasnier, Yannick Teglia","submitted_at":"2026-06-02T15:38:12Z","abstract_excerpt":"Backdoor attacks in Large Language Models (LLMs) are a growing security concern, where models can generate adversary-chosen content. Existing defenses target backdoors one at a time and typically require knowledge of the trigger, leaving the defender at a structural disadvantage when unknown backdoors may exist in a model. We show that backdoor neutralization through unlearning generalizes across backdoors: training a model to ignore a single trigger can also suppress other backdoors that were never explicitly targeted. We study this phenomenon across three model families, whose backdoors were"},"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":"2606.03785","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-06-02T15:38:12Z","cross_cats_sorted":[],"title_canon_sha256":"ce2f13479fec0188afe46200aac6f7d297ba720186dc30f514cd1154b9163437","abstract_canon_sha256":"b7b0c48f68a9776af669bec13f84b5220f023f5fd848f8fad347e3d2a02eef1b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-03T02:06:02.442467Z","signature_b64":"aE3piOj90Hiug//7UAB4wHtdl7eH2HoOp5cRn5rTQ21Vu5ycPFapPm/QZvFoUqkmIFqpb+nPyBD+4gRtFWxzBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"06367c68f8df8c885301c471335e902cdb0346fd9a111517e3aac107a3a831d5","last_reissued_at":"2026-06-03T02:06:02.441999Z","signature_status":"signed_v1","first_computed_at":"2026-06-03T02:06:02.441999Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Backdoor Unlearning Generalization: A Path Toward the Removal of Unknown Triggers in LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Djam\\'e Seddah, Lisa Bouger, Philippe Looubet Moundi, Th\\'eo Lasnier, Yannick Teglia","submitted_at":"2026-06-02T15:38:12Z","abstract_excerpt":"Backdoor attacks in Large Language Models (LLMs) are a growing security concern, where models can generate adversary-chosen content. Existing defenses target backdoors one at a time and typically require knowledge of the trigger, leaving the defender at a structural disadvantage when unknown backdoors may exist in a model. We show that backdoor neutralization through unlearning generalizes across backdoors: training a model to ignore a single trigger can also suppress other backdoors that were never explicitly targeted. We study this phenomenon across three model families, whose backdoors were"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.03785","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/2606.03785/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":"2606.03785","created_at":"2026-06-03T02:06:02.442051+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.03785v1","created_at":"2026-06-03T02:06:02.442051+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.03785","created_at":"2026-06-03T02:06:02.442051+00:00"},{"alias_kind":"pith_short_12","alias_value":"AY3HY2HY36GI","created_at":"2026-06-03T02:06:02.442051+00:00"},{"alias_kind":"pith_short_16","alias_value":"AY3HY2HY36GIQUYB","created_at":"2026-06-03T02:06:02.442051+00:00"},{"alias_kind":"pith_short_8","alias_value":"AY3HY2HY","created_at":"2026-06-03T02:06:02.442051+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/AY3HY2HY36GIQUYBYRYTGXUQFT","json":"https://pith.science/pith/AY3HY2HY36GIQUYBYRYTGXUQFT.json","graph_json":"https://pith.science/api/pith-number/AY3HY2HY36GIQUYBYRYTGXUQFT/graph.json","events_json":"https://pith.science/api/pith-number/AY3HY2HY36GIQUYBYRYTGXUQFT/events.json","paper":"https://pith.science/paper/AY3HY2HY"},"agent_actions":{"view_html":"https://pith.science/pith/AY3HY2HY36GIQUYBYRYTGXUQFT","download_json":"https://pith.science/pith/AY3HY2HY36GIQUYBYRYTGXUQFT.json","view_paper":"https://pith.science/paper/AY3HY2HY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.03785&json=true","fetch_graph":"https://pith.science/api/pith-number/AY3HY2HY36GIQUYBYRYTGXUQFT/graph.json","fetch_events":"https://pith.science/api/pith-number/AY3HY2HY36GIQUYBYRYTGXUQFT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AY3HY2HY36GIQUYBYRYTGXUQFT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AY3HY2HY36GIQUYBYRYTGXUQFT/action/storage_attestation","attest_author":"https://pith.science/pith/AY3HY2HY36GIQUYBYRYTGXUQFT/action/author_attestation","sign_citation":"https://pith.science/pith/AY3HY2HY36GIQUYBYRYTGXUQFT/action/citation_signature","submit_replication":"https://pith.science/pith/AY3HY2HY36GIQUYBYRYTGXUQFT/action/replication_record"}},"created_at":"2026-06-03T02:06:02.442051+00:00","updated_at":"2026-06-03T02:06:02.442051+00:00"}