{"id":"W4385462630","doi":"10.1016/j.artint.2023.103989","title":"Learning reward machines: A study in partially observable reinforcement learning","year":2023,"lang":"en","type":"article","venue":"Artificial Intelligence","topic":"Reinforcement Learning in Robotics","field":"Computer Science","cited_by":14,"is_retracted":false,"has_abstract":false,"ca_institutions":"Toronto Metropolitan University; Vector Institute; University of Toronto","funders":"Fondo Nacional de Desarrollo Científico y Tecnológico; Agencia Nacional de Investigación y Desarrollo; Government of Ontario; Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research; Microsoft Research","keywords":"Reinforcement learning; Observable; Computer science; Set (abstract data type); Artificial intelligence; Task (project management); Function (biology); Optimization problem; Representation (politics); Decomposition; Learning automata; Mathematical optimization; Automaton; Machine learning; Mathematics; Algorithm","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","insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.002094184,0.0002994337,0.0003382525,0.0004684524,0.0004220884,0.0004494473,0.001326954,0.00009308235,0.00009116698],"category_scores_gemma":[0.001060153,0.000310357,0.00009511101,0.002741708,0.00006353985,0.0006937476,0.000802985,0.0008309183,0.002225862],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001365102,"about_ca_system_score_gemma":0.00011429,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0004418079,"about_ca_topic_score_gemma":0.0003059671,"domain_scores_codex":[0.9962094,0.0003721899,0.0009968728,0.0007415404,0.0007887891,0.0008912185],"domain_scores_gemma":[0.9984233,0.000401337,0.0002448406,0.000660705,0.0001281915,0.0001416165],"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":[0.00001829995,0.00006121002,0.03335826,0.00001281145,0.00001684157,0.0001064423,0.007663286,0.9251564,0.0002581961,0.007792035,0.00003468876,0.02552153],"study_design_scores_gemma":[0.00007223813,0.0008037677,0.003760231,0.00004837614,0.000006812252,0.000002636348,0.002867832,0.9884132,0.0009733525,0.001667414,0.001035801,0.0003483799],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.168383,0.00001451818,0.8270178,0.0004129594,0.0006477088,0.000629248,8.705469e-8,0.0009692112,0.001925457],"genre_scores_gemma":[0.9948818,0.0000402415,0.0017606,0.00009517563,0.00008656549,0.0000838156,0.000008110535,0.00003051428,0.003013168],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8264989,"threshold_uncertainty_score":0.9999349,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08641297792112246,"score_gpt":0.3260962182458153,"score_spread":0.2396832403246928,"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."}}