{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:QRR57IYZ7OFTSG36NNSF4ZKKAP","short_pith_number":"pith:QRR57IYZ","schema_version":"1.0","canonical_sha256":"8463dfa319fb8b391b7e6b645e654a03cbaad89a73521b6b9e76152f9c0f2433","source":{"kind":"arxiv","id":"2011.10369","version":3},"attestation_state":"computed","paper":{"title":"ONION: A Simple and Effective Defense Against Textual Backdoor Attacks","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.CY"],"primary_cat":"cs.CL","authors_text":"Fanchao Qi, Maosong Sun, Mukai Li, Yangyi Chen, Yuan Yao, Zhiyuan Liu","submitted_at":"2020-11-20T12:17:21Z","abstract_excerpt":"Backdoor attacks are a kind of emergent training-time threat to deep neural networks (DNNs). They can manipulate the output of DNNs and possess high insidiousness. In the field of natural language processing, some attack methods have been proposed and achieve very high attack success rates on multiple popular models. Nevertheless, there are few studies on defending against textual backdoor attacks. In this paper, we propose a simple and effective textual backdoor defense named ONION, which is based on outlier word detection and, to the best of our knowledge, is the first method that can handle"},"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":"2011.10369","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2020-11-20T12:17:21Z","cross_cats_sorted":["cs.CY"],"title_canon_sha256":"87027a0a725f956957866a6210e0c9a15327968f514fffbf2f2bd7eb22b9ddc9","abstract_canon_sha256":"2d0f2538a55b9f9b7ee6ce538eb919c8a12d4726d887ce597980f1fd302a00bb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:28:50.701023Z","signature_b64":"q7KLzfMHZvdef7eMuVdHRsilBjRq9TANH4TWaiOuHjAV1mtZYlEyPQuXAlGXGrIC0NJjkg98WBiQhoMOvgXvCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8463dfa319fb8b391b7e6b645e654a03cbaad89a73521b6b9e76152f9c0f2433","last_reissued_at":"2026-07-05T03:28:50.700542Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:28:50.700542Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ONION: A Simple and Effective Defense Against Textual Backdoor Attacks","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.CY"],"primary_cat":"cs.CL","authors_text":"Fanchao Qi, Maosong Sun, Mukai Li, Yangyi Chen, Yuan Yao, Zhiyuan Liu","submitted_at":"2020-11-20T12:17:21Z","abstract_excerpt":"Backdoor attacks are a kind of emergent training-time threat to deep neural networks (DNNs). They can manipulate the output of DNNs and possess high insidiousness. In the field of natural language processing, some attack methods have been proposed and achieve very high attack success rates on multiple popular models. Nevertheless, there are few studies on defending against textual backdoor attacks. In this paper, we propose a simple and effective textual backdoor defense named ONION, which is based on outlier word detection and, to the best of our knowledge, is the first method that can handle"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2011.10369","kind":"arxiv","version":3},"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/2011.10369/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":"2011.10369","created_at":"2026-07-05T03:28:50.700595+00:00"},{"alias_kind":"arxiv_version","alias_value":"2011.10369v3","created_at":"2026-07-05T03:28:50.700595+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2011.10369","created_at":"2026-07-05T03:28:50.700595+00:00"},{"alias_kind":"pith_short_12","alias_value":"QRR57IYZ7OFT","created_at":"2026-07-05T03:28:50.700595+00:00"},{"alias_kind":"pith_short_16","alias_value":"QRR57IYZ7OFTSG36","created_at":"2026-07-05T03:28:50.700595+00:00"},{"alias_kind":"pith_short_8","alias_value":"QRR57IYZ","created_at":"2026-07-05T03:28:50.700595+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"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":10,"is_internal_anchor":false},{"citing_arxiv_id":"2504.00446","citing_title":"Exposing the Ghost in the Transformer: Abnormal Detection for Large Language Models via Hidden State Forensics","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2504.05605","citing_title":"ShadowCoT: Cognitive Hijacking for Stealthy Reasoning Backdoors in LLMs","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2512.10998","citing_title":"SCOUT: A Defense Against Data Poisoning Attacks in Fine-Tuned Language Models","ref_index":16,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QRR57IYZ7OFTSG36NNSF4ZKKAP","json":"https://pith.science/pith/QRR57IYZ7OFTSG36NNSF4ZKKAP.json","graph_json":"https://pith.science/api/pith-number/QRR57IYZ7OFTSG36NNSF4ZKKAP/graph.json","events_json":"https://pith.science/api/pith-number/QRR57IYZ7OFTSG36NNSF4ZKKAP/events.json","paper":"https://pith.science/paper/QRR57IYZ"},"agent_actions":{"view_html":"https://pith.science/pith/QRR57IYZ7OFTSG36NNSF4ZKKAP","download_json":"https://pith.science/pith/QRR57IYZ7OFTSG36NNSF4ZKKAP.json","view_paper":"https://pith.science/paper/QRR57IYZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2011.10369&json=true","fetch_graph":"https://pith.science/api/pith-number/QRR57IYZ7OFTSG36NNSF4ZKKAP/graph.json","fetch_events":"https://pith.science/api/pith-number/QRR57IYZ7OFTSG36NNSF4ZKKAP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QRR57IYZ7OFTSG36NNSF4ZKKAP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QRR57IYZ7OFTSG36NNSF4ZKKAP/action/storage_attestation","attest_author":"https://pith.science/pith/QRR57IYZ7OFTSG36NNSF4ZKKAP/action/author_attestation","sign_citation":"https://pith.science/pith/QRR57IYZ7OFTSG36NNSF4ZKKAP/action/citation_signature","submit_replication":"https://pith.science/pith/QRR57IYZ7OFTSG36NNSF4ZKKAP/action/replication_record"}},"created_at":"2026-07-05T03:28:50.700595+00:00","updated_at":"2026-07-05T03:28:50.700595+00:00"}