{"id":"W3212787291","doi":"10.1109/qrs54544.2021.00037","title":"On Assessing The Safety of Reinforcement Learning algorithms Using Formal Methods","year":2021,"lang":"en","type":"article","venue":"2021 IEEE 21st International Conference on Software Quality, Reliability and Security (QRS)","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"Polytechnique Montréal","funders":"","keywords":"Adversarial system; Probabilistic logic; Reinforcement learning; Computer science; Artificial intelligence; Computer security; Multi-agent system; Autonomous agent; Grid; Machine learning; Mathematics","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":[0.02473788,0.001889926,0.001140607,0.002322503,0.001052701,0.002881621,0.001929803,0.001991887,0.002370089],"category_scores_gemma":[0.1348977,0.000672757,0.001550116,0.0009099855,0.005616733,0.0042417,0.002937617,0.003175537,0.0002442992],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00324259,"about_ca_system_score_gemma":0.004552638,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004253694,"about_ca_topic_score_gemma":0.002188022,"domain_scores_codex":[0.9824104,0.01114967,0.0008155123,0.001219123,0.003677322,0.0007280483],"domain_scores_gemma":[0.7167638,0.2581214,0.009750491,0.00799838,0.006019915,0.001346059],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.0001282859,0.0001059601,0.003667109,0.0002106932,0.00007547016,0.00004307006,0.0001937163,0.8959075,0.001190547,0.06793348,0.0002387811,0.03030532],"study_design_scores_gemma":[0.00001836989,0.00007867465,0.0001689808,0.00005214533,0.00001082948,0.00001808368,0.0000215847,0.9593529,0.0007525616,0.03924681,0.0002678457,0.00001126417],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.01673341,0.0002591808,0.9807196,0.0004279727,0.00002835443,0.00008995031,0.00003737808,0.0003483777,0.001355725],"genre_scores_gemma":[0.6192753,0.0006417008,0.3782064,0.0002657871,0.00008968855,0.0004205618,0.000157393,0.0002138935,0.0007292076],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.02473788,"threshold_uncertainty_score":0.130828,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06773412524589584,"score_gpt":0.4105192365666844,"score_spread":0.3427851113207886,"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."}}