{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002000314,0.0008597403,0.0008102857,0.000373727,0.0003649167,0.0004457002,0.001038247,0.001269224,0.001738206],"category_scores_gemma":[0.007186667,0.0003732366,0.0003787995,0.000191651,0.001091731,0.001095961,0.002455837,0.001999835,0.0003106152],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006311057,"about_ca_system_score_gemma":0.0009103137,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001384139,"about_ca_topic_score_gemma":0.001460948,"domain_scores_codex":[0.9989882,0.0003451348,0.00003713324,0.0001628043,0.0002485091,0.0002181834],"domain_scores_gemma":[0.9970394,0.001724149,0.0002343524,0.000530393,0.000311695,0.0001600339],"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":[0.0003329376,0.00008606683,0.001089919,0.00003226946,0.00005353489,0.0001033008,0.00003798126,0.9398679,0.005446353,0.008958172,0.001410243,0.04258137],"study_design_scores_gemma":[0.000006890173,0.00003393639,0.00009138139,0.000003325543,0.000003285621,0.00001293683,0.000003020541,0.9961838,0.001062421,0.002496999,0.00009976514,0.000002272591],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.2299208,0.0004228506,0.756774,0.001045177,0.0001766306,0.00008461227,0.0000833829,0.003297675,0.008194929],"genre_scores_gemma":[0.9780367,0.00003721275,0.02015388,0.0001142968,0.00001230268,0.00002562964,0.00003354772,0.00007095235,0.001515474],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.002000314,"threshold_uncertainty_score":0.01057881,"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."}}