{"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":"codex-gemma-dda1882f352a","candidate_categories":["research_integrity"],"consensus_categories":[],"category_scores_codex":[0.007680622,0.000173457,0.0003079037,0.001429012,0.000805173,0.0003705154,0.00171371,0.00007099303,0.0004823425],"category_scores_gemma":[0.001497566,0.0001667666,0.0001175067,0.002283619,0.00009991034,0.0008412846,0.001028001,0.003419774,0.00003401743],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0007225922,"about_ca_system_score_gemma":0.0004233338,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00008558952,"about_ca_topic_score_gemma":0.00001439829,"domain_scores_codex":[0.9937354,0.00110426,0.001332225,0.0003433353,0.002735107,0.0007496561],"domain_scores_gemma":[0.9975918,0.0005915556,0.0005737434,0.0004105511,0.0006643192,0.000168032],"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":[0.0001468123,0.00003570172,0.0008341765,0.0000192612,0.00002008649,0.0001063135,0.002774453,0.9203317,0.003979521,0.01172592,0.0001459679,0.05988015],"study_design_scores_gemma":[0.0001137608,0.002815373,0.0001975151,0.0001026192,0.000006892068,0.0001379363,0.007290337,0.9629144,0.004042804,0.01653883,0.00557523,0.0002643098],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.06806064,0.0001578907,0.9287345,0.001028922,0.001010319,0.0002438727,3.459332e-7,0.00003746507,0.0007260342],"genre_scores_gemma":[0.9942685,0.00007585859,0.004872578,0.00007221857,0.0002984882,0.00001080453,0.000002357972,0.00002001484,0.0003791165],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9262079,"threshold_uncertainty_score":0.9988794,"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."}}