{"id":"W3176864543","doi":"10.65109/rgis8995","title":"Transferable Environment Poisoning: Training-time Attack on Reinforcement Learning","year":2021,"lang":"en","type":"article","venue":"","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":8,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Reinforcement; Training (meteorology); Reinforcement learning; Computer science; Computer security; Psychology; Artificial intelligence; Social psychology; Geography; Meteorology","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001708486,0.0008245305,0.0006683102,0.0003264006,0.0003604644,0.0007084072,0.001010896,0.001155233,0.001587976],"category_scores_gemma":[0.007616469,0.0002356754,0.0005677841,0.0002045274,0.00200574,0.001489966,0.002231629,0.002127249,0.0002556315],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0007682688,"about_ca_system_score_gemma":0.0007948016,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0009366955,"about_ca_topic_score_gemma":0.0006286719,"domain_scores_codex":[0.9987327,0.000524169,0.00004991346,0.0002240408,0.0002807936,0.0001884289],"domain_scores_gemma":[0.9962739,0.002254725,0.0004153853,0.0007072377,0.0001748872,0.0001738369],"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":[0.0001824884,0.0001091031,0.001195438,0.00005603308,0.00007060882,0.0002595702,0.0001334065,0.9080709,0.009294529,0.04806574,0.0008597875,0.03170235],"study_design_scores_gemma":[0.00001274718,0.00006308247,0.00009327297,0.00000513919,0.000004708409,0.00003599548,0.000007957972,0.982132,0.001980955,0.01533299,0.0003259405,0.000005279946],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.05315435,0.0001349826,0.9416817,0.000398912,0.00004480456,0.00008903255,0.00002621569,0.0008569657,0.003612934],"genre_scores_gemma":[0.954455,0.00008294314,0.04350144,0.0001659199,0.00002165338,0.00007979348,0.00002232431,0.00005916304,0.001611711],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.001708486,"threshold_uncertainty_score":0.009035468,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03446290798627397,"score_gpt":0.2597096128112343,"score_spread":0.2252467048249604,"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."}}