{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:IK3D43V5RZJZUM57GO7ECNPDB3","short_pith_number":"pith:IK3D43V5","schema_version":"1.0","canonical_sha256":"42b63e6ebd8e539a33bf33be4135e30eeb6e17cc8a4c7bd393176126e809ca34","source":{"kind":"arxiv","id":"2403.17155","version":1},"attestation_state":"computed","paper":{"title":"Task-Agnostic Detector for Insertion-Based Backdoor Attacks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.CL","authors_text":"Chao Chen, Haibin Ling, Lu Pang, Songzhu Zheng, Susmit Jha, Weimin Lyu, Xiao Lin","submitted_at":"2024-03-25T20:12:02Z","abstract_excerpt":"Textual backdoor attacks pose significant security threats. Current detection approaches, typically relying on intermediate feature representation or reconstructing potential triggers, are task-specific and less effective beyond sentence classification, struggling with tasks like question answering and named entity recognition. We introduce TABDet (Task-Agnostic Backdoor Detector), a pioneering task-agnostic method for backdoor detection. TABDet leverages final layer logits combined with an efficient pooling technique, enabling unified logit representation across three prominent NLP tasks. TAB"},"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":"2403.17155","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-03-25T20:12:02Z","cross_cats_sorted":["cs.CR"],"title_canon_sha256":"915787e96814d68795fee4c2fd18cf03cf9156b497c7c283562d28bd23bd6bda","abstract_canon_sha256":"1d6729da7ff80bf48e76e6f1298c89b7b26074b99b076b62007182cffd5426c5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:00:46.014268Z","signature_b64":"7Kfy6H1KnMSZJpOjcBZ4BqSUE2qkz+U5ALr14O3NKVAzJQ6ltZUdbVRmEs3ZyABRkHxbpmI8aQD98lnygsbfCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"42b63e6ebd8e539a33bf33be4135e30eeb6e17cc8a4c7bd393176126e809ca34","last_reissued_at":"2026-07-05T08:00:46.013814Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:00:46.013814Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Task-Agnostic Detector for Insertion-Based Backdoor Attacks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.CL","authors_text":"Chao Chen, Haibin Ling, Lu Pang, Songzhu Zheng, Susmit Jha, Weimin Lyu, Xiao Lin","submitted_at":"2024-03-25T20:12:02Z","abstract_excerpt":"Textual backdoor attacks pose significant security threats. Current detection approaches, typically relying on intermediate feature representation or reconstructing potential triggers, are task-specific and less effective beyond sentence classification, struggling with tasks like question answering and named entity recognition. We introduce TABDet (Task-Agnostic Backdoor Detector), a pioneering task-agnostic method for backdoor detection. TABDet leverages final layer logits combined with an efficient pooling technique, enabling unified logit representation across three prominent NLP tasks. TAB"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.17155","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/2403.17155/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":"2403.17155","created_at":"2026-07-05T08:00:46.013872+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.17155v1","created_at":"2026-07-05T08:00:46.013872+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.17155","created_at":"2026-07-05T08:00:46.013872+00:00"},{"alias_kind":"pith_short_12","alias_value":"IK3D43V5RZJZ","created_at":"2026-07-05T08:00:46.013872+00:00"},{"alias_kind":"pith_short_16","alias_value":"IK3D43V5RZJZUM57","created_at":"2026-07-05T08:00:46.013872+00:00"},{"alias_kind":"pith_short_8","alias_value":"IK3D43V5","created_at":"2026-07-05T08:00:46.013872+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.31309","citing_title":"CSO-LLM: Class Subspace Orthogonalization for Post-Training Backdoor Detection and Trigger Inversion in LLMs","ref_index":8,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IK3D43V5RZJZUM57GO7ECNPDB3","json":"https://pith.science/pith/IK3D43V5RZJZUM57GO7ECNPDB3.json","graph_json":"https://pith.science/api/pith-number/IK3D43V5RZJZUM57GO7ECNPDB3/graph.json","events_json":"https://pith.science/api/pith-number/IK3D43V5RZJZUM57GO7ECNPDB3/events.json","paper":"https://pith.science/paper/IK3D43V5"},"agent_actions":{"view_html":"https://pith.science/pith/IK3D43V5RZJZUM57GO7ECNPDB3","download_json":"https://pith.science/pith/IK3D43V5RZJZUM57GO7ECNPDB3.json","view_paper":"https://pith.science/paper/IK3D43V5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.17155&json=true","fetch_graph":"https://pith.science/api/pith-number/IK3D43V5RZJZUM57GO7ECNPDB3/graph.json","fetch_events":"https://pith.science/api/pith-number/IK3D43V5RZJZUM57GO7ECNPDB3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IK3D43V5RZJZUM57GO7ECNPDB3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IK3D43V5RZJZUM57GO7ECNPDB3/action/storage_attestation","attest_author":"https://pith.science/pith/IK3D43V5RZJZUM57GO7ECNPDB3/action/author_attestation","sign_citation":"https://pith.science/pith/IK3D43V5RZJZUM57GO7ECNPDB3/action/citation_signature","submit_replication":"https://pith.science/pith/IK3D43V5RZJZUM57GO7ECNPDB3/action/replication_record"}},"created_at":"2026-07-05T08:00:46.013872+00:00","updated_at":"2026-07-05T08:00:46.013872+00:00"}