{"id":"W6893889155","doi":"10.5281/zenodo.6464098","title":"FalsifAI: Falsification of AI-Enabled Hybrid Control Systems Guided by Time-Aware Coverage Criteria","year":2022,"lang":"en","type":"article","venue":"Zenodo (CERN European Organization for Nuclear Research)","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"","keywords":"Robustness (evolution); Artificial neural network; Hybrid system; Exploit; Context (archaeology); Control system; Controller (irrigation)","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":[0.003186662,0.00162803,0.001092616,0.001415293,0.0007466577,0.002145164,0.001603762,0.001537086,0.003132376],"category_scores_gemma":[0.02068595,0.0005150801,0.001754508,0.0004853214,0.003221319,0.002143094,0.002751251,0.001814847,0.0002995244],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001600388,"about_ca_system_score_gemma":0.001816416,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003737598,"about_ca_topic_score_gemma":0.002517061,"domain_scores_codex":[0.9967468,0.0008100116,0.0002092468,0.0005855471,0.001187644,0.0004608228],"domain_scores_gemma":[0.9862393,0.009961108,0.001100753,0.001033727,0.001321026,0.0003441256],"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":[0.000434731,0.00007840044,0.003990275,0.0003674981,0.0001437113,0.0009795595,0.0004987931,0.8174577,0.01434551,0.1243179,0.001479528,0.03590639],"study_design_scores_gemma":[0.00002508326,0.00007523935,0.0002097272,0.00004647579,0.00002265924,0.00009425841,0.00004047359,0.946556,0.004567338,0.04754626,0.0007964143,0.00002014805],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.04703277,0.0003598101,0.9459962,0.0004408775,0.00007782219,0.0001206472,0.0001955103,0.00143563,0.004340793],"genre_scores_gemma":[0.8993075,0.0002379118,0.0977923,0.0002384688,0.00005880239,0.0002206696,0.0003205486,0.000284197,0.00153959],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.003737598,"threshold_uncertainty_score":0.01685286,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01905003428521876,"score_gpt":0.2506281898350346,"score_spread":0.2315781555498158,"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."}}