{"id":"W4385060545","doi":"10.5220/0012139900003555","title":"Multi-Environment Training Against Reward Poisoning Attacks on Deep Reinforcement Learning","year":2023,"lang":"en","type":"article","venue":"","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":false,"ca_institutions":"Université du Québec en Outaouais","funders":"","keywords":"Reinforcement learning; Computer science; Training (meteorology); Reinforcement; Artificial intelligence; Psychology; Social psychology","routes":{"ca_aff":true,"ca_fund":false,"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":[0.001010969,0.0003125168,0.0002818776,0.0003088147,0.0006127715,0.0001969181,0.0009517894,0.00011027,0.0001033734],"category_scores_gemma":[0.000348206,0.0002996528,0.0001283005,0.0005845056,0.00005126539,0.0004660146,0.000827163,0.000737846,0.001372096],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000209803,"about_ca_system_score_gemma":0.00004104015,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00001891337,"about_ca_topic_score_gemma":0.000002233411,"domain_scores_codex":[0.9971425,0.0001849048,0.0004308544,0.0007627422,0.0006879528,0.0007910455],"domain_scores_gemma":[0.9986576,0.0003138742,0.0001906601,0.0006331252,0.00002381717,0.0001809583],"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":[0.000006011703,0.00001355158,0.0007230263,0.000007041144,0.00002289116,0.00005644265,0.003576746,0.9357802,0.0003134598,0.002699664,0.0001203181,0.05668059],"study_design_scores_gemma":[0.0006857956,0.0001360175,0.001506804,0.00005208281,0.000005007124,0.000003562713,0.001070057,0.9885162,0.0001420867,0.00005586209,0.007454481,0.0003720099],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.008833831,0.00001282747,0.9753905,0.0009995945,0.0004569902,0.0002209237,9.436237e-8,0.001379525,0.0127057],"genre_scores_gemma":[0.8597837,0.00003980125,0.1338941,0.000834941,0.0001357904,0.00003514123,0.00001719549,0.00004535786,0.00521401],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.8509499,"threshold_uncertainty_score":0.9999456,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04147161911139761,"score_gpt":0.2807891303442872,"score_spread":0.2393175112328896,"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."}}