{"id":"W4390872800","doi":"10.1109/iccv51070.2023.00190","title":"SAFARI: Versatile and Efficient Evaluations for Robustness of Interpretability","year":2023,"lang":"en","type":"article","venue":"","topic":"Explainable Artificial Intelligence (XAI)","field":"Computer Science","cited_by":23,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"University of Warwick; Transport Canada","keywords":"Interpretability; Robustness (evolution); Computer science; Machine learning; Artificial intelligence; Probabilistic logic; Bespoke; Data mining; Trustworthiness","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":[0.01987899,0.003358389,0.001964418,0.004159247,0.0009083607,0.003763842,0.004087392,0.004123673,0.00399778],"category_scores_gemma":[0.09437552,0.0008236407,0.001827862,0.001831224,0.002423448,0.003930952,0.004082917,0.004578386,0.001103944],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002128951,"about_ca_system_score_gemma":0.002356058,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002492202,"about_ca_topic_score_gemma":0.002870679,"domain_scores_codex":[0.9863259,0.006395836,0.001083392,0.001392587,0.004302506,0.000499856],"domain_scores_gemma":[0.9462819,0.03991575,0.002712079,0.006206512,0.0041579,0.000725928],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.001045757,0.0003566531,0.008662126,0.001218415,0.000748701,0.0002517971,0.000291918,0.6478149,0.009846132,0.0350405,0.01268758,0.2820355],"study_design_scores_gemma":[0.00004400443,0.0001462845,0.0005502592,0.00007618474,0.00004006102,0.00006823864,0.0000381391,0.9760232,0.004184902,0.01725035,0.001550369,0.00002807348],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.03441363,0.002210259,0.9510336,0.0007218688,0.0002351638,0.0002868191,0.0007147125,0.006710688,0.003673249],"genre_scores_gemma":[0.3700435,0.0006627685,0.6219408,0.0005430199,0.0001833747,0.0006562587,0.002425588,0.002021581,0.001523102],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01987899,"threshold_uncertainty_score":0.1051314,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04740619807341593,"score_gpt":0.3414541511880035,"score_spread":0.2940479531145875,"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."}}