{"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":"codex-gemma-dda1882f352a","candidate_categories":["sts","insufficient_payload"],"consensus_categories":["insufficient_payload"],"category_scores_codex":[0.001577985,0.0001663863,0.0002574739,0.0002372383,0.002316502,0.0007604487,0.002471053,0.00003555415,0.004306466],"category_scores_gemma":[0.0005397512,0.0001930298,0.00006308805,0.0006594579,0.00008179536,0.0005747399,0.002058092,0.0004117644,0.0008851393],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002516664,"about_ca_system_score_gemma":0.00001241604,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00004844203,"about_ca_topic_score_gemma":2.053384e-8,"domain_scores_codex":[0.9964705,0.001445873,0.0004586258,0.0005453562,0.0007127652,0.0003669227],"domain_scores_gemma":[0.9980898,0.00006942746,0.0003120218,0.0008098097,0.0005931972,0.0001257656],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.0001379306,0.0004481571,0.00001583095,0.0001493732,0.0001164474,0.00004848711,0.001318778,0.07440709,0.04907789,0.01782183,0.83616,0.02029817],"study_design_scores_gemma":[0.001319917,0.0002874286,0.00006529525,0.00001654015,0.00001363037,0.0001267124,0.000114816,0.3699864,0.0004940001,0.0002295797,0.6271067,0.00023889],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.008149588,0.00009365731,0.9743445,0.001231392,0.0003314024,0.0007837334,0.0005298235,0.0010292,0.0135067],"genre_scores_gemma":[0.9964805,0.000007767327,0.0003598378,0.0002098522,0.00006861887,3.55873e-7,0.001038737,0.0005226305,0.001311704],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9883309,"threshold_uncertainty_score":0.9998928,"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."}}