{"id":"W3020277389","doi":"10.48550/arxiv.2004.12399","title":"Reinforcement Learning Generalization with Surprise Minimization","year":2020,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Reinforcement Learning in Robotics","field":"Computer Science","cited_by":8,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Surprise; Generalization; Computer science; Reinforcement learning; Artificial intelligence; Robustness (evolution); Machine learning; Mathematics; Psychology","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0001496604,0.0003550188,0.0003049305,0.0002279876,0.0002243991,0.0002630737,0.001279897,0.0002219556,0.00004043587],"category_scores_gemma":[0.00006639107,0.000393093,0.0001069867,0.0007883495,0.00006399227,0.0005063337,0.001492015,0.0006480915,0.00008665998],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002601314,"about_ca_system_score_gemma":0.0002509943,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00005172878,"about_ca_topic_score_gemma":0.000005017397,"domain_scores_codex":[0.9979971,0.0001734941,0.0002751375,0.000993243,0.0002207693,0.0003402264],"domain_scores_gemma":[0.998143,0.00005328056,0.0005278075,0.0008366317,0.0002555005,0.0001838113],"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":[0.00002591576,0.00001068816,0.002901306,0.00007016387,0.0000732688,0.00009716191,0.0004246715,0.9524704,0.00001366634,0.04360986,0.0002130774,0.00008988108],"study_design_scores_gemma":[0.00048646,0.0001638268,0.0002812721,0.00008966636,0.00006616267,0.000002333551,0.00004365838,0.9964676,0.0001025457,0.0005241982,0.001313433,0.000458864],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.005995169,0.00001321482,0.9891825,0.0001740973,0.0002857024,0.0003743797,6.003239e-7,0.0005139561,0.003460377],"genre_scores_gemma":[0.985941,0.0001683749,0.009273266,0.0001620204,0.00007835034,0.000001490706,0.0001354734,0.0000329693,0.004207043],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9799458,"threshold_uncertainty_score":0.9998521,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06140952338521455,"score_gpt":0.1835702624956942,"score_spread":0.1221607391104796,"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."}}