{"id":"W4404102001","doi":"10.1109/tse.2024.3491496","title":"SMARLA: A Safety Monitoring Approach for Deep Reinforcement Learning Agents","year":2024,"lang":"en","type":"article","venue":"IEEE Transactions on Software Engineering","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":7,"is_retracted":false,"has_abstract":true,"ca_institutions":"École de Technologie Supérieure; University of Ottawa","funders":"Natural Sciences and Engineering Research Council of Canada; Science Foundation Ireland","keywords":"Computer science; Reinforcement learning; Artificial intelligence; Machine learning; Software engineering","routes":{"ca_aff":true,"ca_fund":true,"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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.00032799,0.0002786597,0.0002112844,0.0003203269,0.0003408185,0.0002479651,0.0004670561,0.0001122113,0.00001403571],"category_scores_gemma":[0.00006394016,0.0003007189,0.0001988654,0.0005472695,0.00001259863,0.0005918661,0.000009684108,0.0007181081,0.00001790137],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002904454,"about_ca_system_score_gemma":0.00004277338,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000008319195,"about_ca_topic_score_gemma":2.033895e-7,"domain_scores_codex":[0.998347,0.00002815091,0.000311574,0.0005350021,0.0003288645,0.0004494149],"domain_scores_gemma":[0.9990811,0.00037901,0.00003778033,0.0003388387,0.00004698621,0.0001162803],"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":[0.00001000533,0.00001591549,0.00001958121,0.000165035,0.00008058507,0.000008184721,0.0006327659,0.9564568,0.00008237822,0.0004299252,0.000007503228,0.04209126],"study_design_scores_gemma":[0.0002674853,0.0000718774,0.00003854035,0.0001396912,0.00002917573,0.00001362709,0.00003347708,0.9955176,0.001428522,0.000008901276,0.002150449,0.0003007243],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.0002550437,0.0001313798,0.9943503,0.00002397268,0.002775432,0.0003278845,0.000001447451,0.002061277,0.00007325703],"genre_scores_gemma":[0.6131497,0.00002035253,0.3860597,0.000009637735,0.0001684929,0.0001408016,0.000003187169,0.00005144694,0.0003967038],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.6128947,"threshold_uncertainty_score":0.9999445,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01605624727510781,"score_gpt":0.2502795647694263,"score_spread":0.2342233174943184,"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."}}