{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001581463,0.0003633626,0.0003434367,0.0002494132,0.0002852192,0.001231364,0.0007383389,0.001070848,0.001306084],"category_scores_gemma":[0.009810132,0.0003142186,0.0002912284,0.0001446363,0.001487981,0.002089791,0.000613134,0.001013908,0.0005271334],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0007888103,"about_ca_system_score_gemma":0.0007537205,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004048893,"about_ca_topic_score_gemma":0.002545191,"domain_scores_codex":[0.9989572,0.0005132261,0.00004506767,0.0001805431,0.0001756866,0.0001284527],"domain_scores_gemma":[0.9975187,0.001174954,0.0004035014,0.000446326,0.0002981262,0.0001585068],"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":[0.0001780568,0.00008799494,0.00901522,0.0001187101,0.0001372098,0.0002990118,0.0006037204,0.8582607,0.01152028,0.06899774,0.002206177,0.04857528],"study_design_scores_gemma":[0.00003278,0.00004939894,0.001442718,0.00002647907,0.0000220501,0.00007475511,0.0001126617,0.9375544,0.002449414,0.05658626,0.001619436,0.00002972135],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3442026,0.0005377508,0.6317413,0.00343971,0.000162759,0.0001158385,0.0001457882,0.001506246,0.01814788],"genre_scores_gemma":[0.9723117,0.0001955017,0.02473112,0.0001939832,0.00001605349,0.00004567671,0.00004298084,0.0001042043,0.002358739],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.004048893,"threshold_uncertainty_score":0.008363664,"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."}}