{"id":"W3216764023","doi":"10.48550/arxiv.2110.03184","title":"Explaining Deep Reinforcement Learning Agents In The Atari Domain through a Surrogate Model","year":2021,"lang":"en","type":"article","venue":"arXiv (Cornell University)","topic":"Explainable Artificial Intelligence (XAI)","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"","keywords":"Reinforcement learning; Computer science; Artificial intelligence; Suite; Domain (mathematical analysis); Transformation (genetics); Replicate; Deep learning; Representation (politics); Deep neural networks; Surrogate model; Machine learning; Mathematics; Law","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":[],"consensus_categories":[],"category_scores_codex":[0.0006051956,0.0001872611,0.0001776178,0.0001045755,0.0004016237,0.0001694679,0.001293402,0.00007640709,0.00004550925],"category_scores_gemma":[0.00009963464,0.0001867943,0.000103684,0.001438799,0.00007503358,0.001287771,0.000564081,0.0003638699,0.0001473136],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001982839,"about_ca_system_score_gemma":0.0001333041,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002158262,"about_ca_topic_score_gemma":0.0003958869,"domain_scores_codex":[0.9980841,0.0003054798,0.0002384189,0.0006616091,0.0001740389,0.0005363961],"domain_scores_gemma":[0.9987123,0.0002116994,0.000115072,0.0007710405,0.0001138588,0.00007601589],"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":[0.000008050812,0.00003343917,0.0007287117,0.000004668279,0.00000953374,0.0007316882,0.006646645,0.6283864,0.00006840319,0.3631999,0.0000320331,0.0001505253],"study_design_scores_gemma":[0.000248402,0.00004084252,0.00009518681,0.00002634427,0.000007900359,0.00001091424,0.00664584,0.9402797,0.001037929,0.05065161,0.0007426954,0.0002126835],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.2554691,0.00002655531,0.7336421,0.0002341372,0.00008573745,0.0001190605,2.142276e-7,0.00007305153,0.01035009],"genre_scores_gemma":[0.9942412,0.00008634159,0.003876179,0.0005915459,0.00002104023,0.000001888231,0.000005469357,0.00001028329,0.001165999],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.7387722,"threshold_uncertainty_score":0.761725,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1407612288090627,"score_gpt":0.2314626391637081,"score_spread":0.0907014103546454,"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."}}