{"id":"W4404780969","doi":"10.18653/v1/2024.findings-emnlp.990","title":"Improving Adversarial Robustness in Vision-Language Models with Architecture and Prompt Design","year":2024,"lang":"en","type":"article","venue":"","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":false,"ca_institutions":"","funders":"Alliance de recherche numérique du Canada; Canada Excellence Research Chairs, Government of Canada; Canadian Institute for Advanced Research","keywords":"Robustness (evolution); Computer science; Adversarial system; Architecture; Artificial intelligence; Computer architecture","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002725976,0.001477584,0.001228363,0.0005545098,0.0005878735,0.00113688,0.00194394,0.002303214,0.003661445],"category_scores_gemma":[0.01423367,0.0009058258,0.0008811669,0.0004015748,0.001438851,0.002792106,0.003489297,0.003689044,0.001176813],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009909671,"about_ca_system_score_gemma":0.001620776,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002454766,"about_ca_topic_score_gemma":0.002885178,"domain_scores_codex":[0.9986479,0.0004752011,0.00006653932,0.0003282387,0.0003027097,0.0001793749],"domain_scores_gemma":[0.9962893,0.002276185,0.0002271689,0.0005411287,0.0005038257,0.0001624349],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"not_applicable","study_design_scores_codex":[0.0002888544,0.0001067444,0.0005485655,0.0001237551,0.00007526024,0.0001106415,0.00009753693,0.8613207,0.01345881,0.03307999,0.003051291,0.08773783],"study_design_scores_gemma":[0.000008417933,0.00002996292,0.00004276227,0.000005846077,0.000009253424,0.00001902388,0.000004154525,0.9874077,0.002330421,0.009852074,0.0002836773,0.000006803216],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.008188864,0.0001678787,0.9891807,0.0003328118,0.00005945135,0.00003118751,0.00003825965,0.000979441,0.001021309],"genre_scores_gemma":[0.7688703,0.0003135047,0.2223499,0.0006848017,0.000140248,0.0001913436,0.0002607202,0.0005694265,0.006619796],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.003661445,"threshold_uncertainty_score":0.01441652,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.008985468768117039,"score_gpt":0.2443913261637576,"score_spread":0.2354058573956406,"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."}}