{"id":"W3173549437","doi":"10.1609/aaai.v35i13.17378","title":"How RL Agents Behave When Their Actions Are Modified","year":2021,"lang":"en","type":"article","venue":"","topic":"Reinforcement Learning in Robotics","field":"Computer Science","cited_by":10,"is_retracted":false,"has_abstract":true,"ca_institutions":"Vector Institute; University of Toronto","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Reinforcement learning; Action (physics); Supervisor; Markov decision process; Computer science; Process (computing); Risk analysis (engineering); Reinforcement; Control (management); Intervention (counseling); Artificial intelligence; Q-learning; Machine learning; Markov process; Psychology; Social psychology; Business; Economics; Mathematics","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":[],"consensus_categories":[],"category_scores_codex":[0.00006834111,0.0001120914,0.000111542,0.00005042958,0.000151536,0.0005365766,0.0005446892,0.0000546784,0.0001153479],"category_scores_gemma":[0.00005825835,0.00009655443,0.0000730336,0.0001920148,0.00001984728,0.0004705749,0.0003577466,0.000150398,0.000117679],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00004479664,"about_ca_system_score_gemma":0.00005628879,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000008233896,"about_ca_topic_score_gemma":0.00001029043,"domain_scores_codex":[0.9990732,0.0000471574,0.0001150224,0.0002849151,0.0002442464,0.0002354428],"domain_scores_gemma":[0.9989638,0.00005135052,0.000072738,0.0007014385,0.0001263124,0.00008433627],"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":[0.00000371713,0.0002011252,0.004175682,0.00004265894,0.0002070443,0.0001879202,0.004017468,0.7230121,0.003011163,0.1875754,0.06084736,0.01671829],"study_design_scores_gemma":[0.0003189582,0.000043359,0.004632968,0.0000229012,0.00001077329,0.00002415375,0.0006068277,0.929354,0.007039158,0.001126993,0.05653165,0.0002882962],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.001617706,0.00001660186,0.9591709,0.005132262,0.0004631737,0.00007225038,8.890188e-7,0.0002160268,0.03331013],"genre_scores_gemma":[0.8201033,0.00001590363,0.05330507,0.0009189518,0.00006686767,0.0000108679,0.000008459774,0.00001165273,0.1255589],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9058659,"threshold_uncertainty_score":0.517422,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0900760941624977,"score_gpt":0.270716654948072,"score_spread":0.1806405607855743,"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."}}