{"id":"W4389520058","doi":"10.18653/v1/2023.findings-emnlp.35","title":"Toward Stronger Textual Attack Detectors","year":2023,"lang":"en","type":"article","venue":"","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Grand Équipement National De Calcul Intensif","keywords":"Adversarial system; Computer science; Benchmark (surveying); Hyperparameter; Artificial intelligence; Machine learning; Trustworthiness; 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":["insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.0002717579,0.0001162633,0.0001144857,0.0001345832,0.0001081374,0.0001103653,0.0008847149,0.00005347285,0.0001877197],"category_scores_gemma":[0.0001364325,0.0001009307,0.00005738273,0.0007646651,0.00003172396,0.0003726128,0.000628037,0.0001914705,0.001873125],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003252618,"about_ca_system_score_gemma":0.00003963333,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002959964,"about_ca_topic_score_gemma":0.000007090867,"domain_scores_codex":[0.998815,0.00005779882,0.000143231,0.0003398882,0.0002910425,0.0003530546],"domain_scores_gemma":[0.9992837,0.0001496696,0.00003878134,0.0004126538,0.0000315046,0.00008370859],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0000184156,0.00006873491,0.01803138,0.00005230883,0.0001132306,0.0003523559,0.004405257,0.295782,0.001490608,0.2747159,0.06512006,0.3398498],"study_design_scores_gemma":[0.0004402379,0.00008130261,0.01683426,0.00001007733,0.000005487902,0.00001249557,0.0003488674,0.9574885,0.0008087677,0.001022877,0.02253038,0.0004167574],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.04861455,0.000007487018,0.9290543,0.001342465,0.0006002392,0.00008588036,5.194379e-7,0.00163895,0.01865557],"genre_scores_gemma":[0.9623039,0.000001838659,0.03348445,0.0001553639,0.0001400672,0.000006448465,0.000002017609,0.00001271912,0.003893236],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9136893,"threshold_uncertainty_score":0.998904,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.051490275223493,"score_gpt":0.3023379201925875,"score_spread":0.2508476449690946,"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."}}