{"id":"W4385572282","doi":"10.18653/v1/2023.findings-acl.496","title":"Impact of Adversarial Training on Robustness and Generalizability of Language Models","year":2023,"lang":"en","type":"article","venue":"","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":7,"is_retracted":false,"has_abstract":true,"ca_institutions":"Dalhousie University","funders":"","keywords":"Adversarial system; Generalizability theory; Computer science; Embedding; Robustness (evolution); Artificial intelligence; Generalization; Transformer; Machine learning; Training set; Mathematics; Engineering","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.007468306,0.001631001,0.0009751957,0.0007424771,0.0006001582,0.001373394,0.001323543,0.001516232,0.002195161],"category_scores_gemma":[0.05145615,0.0006006792,0.001018222,0.0003714966,0.003147094,0.002677034,0.003752938,0.003653392,0.0004103436],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001147654,"about_ca_system_score_gemma":0.000789498,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001856847,"about_ca_topic_score_gemma":0.001324125,"domain_scores_codex":[0.9968836,0.001495652,0.0001660605,0.0005036208,0.000591461,0.0003596395],"domain_scores_gemma":[0.9626355,0.02955766,0.00163782,0.004812849,0.0008895398,0.0004665834],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0002351702,0.00005251552,0.002024976,0.0001494196,0.0001221495,0.0001615615,0.0001432414,0.9527508,0.006793004,0.01510488,0.0005826873,0.02187947],"study_design_scores_gemma":[0.00001435988,0.0002399066,0.001084349,0.00007498381,0.00003309673,0.0001734643,0.0000683555,0.9645668,0.007093055,0.0260088,0.0006192321,0.00002365308],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.2738664,0.002007945,0.705951,0.0025059,0.0001880983,0.0001852899,0.0003286223,0.001692781,0.01327393],"genre_scores_gemma":[0.974256,0.0005978733,0.02297056,0.0003038935,0.00005526864,0.00008391532,0.0001892434,0.0002430552,0.00130017],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.007468306,"threshold_uncertainty_score":0.03949666,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04219023531837281,"score_gpt":0.3193435283298187,"score_spread":0.2771532930114459,"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."}}