{"id":"W3092156990","doi":"10.1613/jair.1.12440","title":"Reward Machines: Exploiting Reward Function Structure in Reinforcement Learning","year":2022,"lang":"en","type":"article","venue":"Journal of Artificial Intelligence Research","topic":"Reinforcement Learning in Robotics","field":"Computer Science","cited_by":172,"is_retracted":false,"has_abstract":true,"ca_institutions":"Vector Institute; University of Toronto","funders":"Natural Sciences and Engineering Research Council of Canada; University of Toronto; Government of Canada; Canadian Institute for Advanced Research; Vector Institute; Microsoft Research","keywords":"Reinforcement learning; Computer science; Exploit; Counterfactual thinking; Function (biology); Artificial intelligence; Finite-state machine; Machine learning; Algorithm; Psychology","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002877827,0.0008595068,0.0008421073,0.0005075395,0.0004214588,0.001370213,0.001430926,0.001263346,0.003297319],"category_scores_gemma":[0.01116771,0.0005888832,0.0008212834,0.0004891472,0.002738575,0.002801512,0.001508486,0.002715355,0.000561111],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001035382,"about_ca_system_score_gemma":0.001184069,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001448808,"about_ca_topic_score_gemma":0.001353948,"domain_scores_codex":[0.9983287,0.0009114333,0.00007595006,0.0002666799,0.0003147154,0.0001025888],"domain_scores_gemma":[0.9943669,0.004296996,0.0003662179,0.0006042265,0.0002319131,0.0001337995],"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":[0.0002764249,0.0001254525,0.001026875,0.0001839499,0.00006096576,0.0001488147,0.0002818138,0.6184499,0.005538596,0.2792026,0.001410014,0.09329476],"study_design_scores_gemma":[0.00002804241,0.00005976033,0.00006505713,0.00002120178,0.000009422978,0.00002267068,0.000007260441,0.896973,0.002085738,0.09931213,0.001402336,0.00001328738],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.004660038,0.00008334038,0.9937045,0.0001250549,0.00001676991,0.00003502243,0.00002529977,0.0005355071,0.0008144121],"genre_scores_gemma":[0.4603901,0.0002563169,0.5362601,0.0002303978,0.00005716171,0.0003757845,0.0001056874,0.0002951884,0.002029386],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.003297319,"threshold_uncertainty_score":0.01521963,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1113564636583909,"score_gpt":0.3745631904237698,"score_spread":0.2632067267653789,"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."}}