{"id":"W3175052694","doi":"10.18653/v1/2021.acl-long.377","title":"Turn the Combination Lock: Learnable Textual Backdoor Attacks via Word Substitution","year":2021,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":93,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"National Key Research and Development Program of China","keywords":"Backdoor; Substitution (logic); Computer science; Lock (firearm); Word (group theory); Computational linguistics; Natural language processing; Artificial intelligence; Linguistics; History; Programming language; Philosophy; Computer security","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0002693063,0.00008282212,0.00008141359,0.00002935894,0.0001958894,0.0002148029,0.00046518,0.00004972812,0.0001559173],"category_scores_gemma":[0.00004909183,0.00006017481,0.00004354004,0.0003196767,0.00002991363,0.0004127946,0.0002210955,0.0001499257,0.0002267047],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005214403,"about_ca_system_score_gemma":0.00009762691,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00006326516,"about_ca_topic_score_gemma":0.00006828077,"domain_scores_codex":[0.9990127,0.00007493886,0.0001636521,0.0002880986,0.0002580261,0.0002025743],"domain_scores_gemma":[0.9991717,0.00006579619,0.00004326227,0.0005336754,0.0001426792,0.00004291836],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000003508878,0.0001356349,0.00112562,0.0000146739,0.00002230633,0.00003546779,0.0006699922,0.008337239,0.003870335,0.8376976,0.00331321,0.1447744],"study_design_scores_gemma":[0.0004556337,0.00002526534,0.002864627,0.00001654475,0.000007111224,0.00006633624,0.0001534973,0.9432087,0.01438073,0.01220731,0.02640144,0.0002128307],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.03447988,0.000112716,0.9481348,0.00441245,0.0003995999,0.00008403338,1.876254e-7,0.0001255827,0.01225077],"genre_scores_gemma":[0.971921,0.000009042103,0.01920835,0.0006571712,0.00007454155,0.000005884727,0.000003473842,0.000004182651,0.00811641],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9374411,"threshold_uncertainty_score":0.2913906,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02142620390299724,"score_gpt":0.2467543115358606,"score_spread":0.2253281076328634,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}