{"id":"W7126269717","doi":"10.21428/594757db.25036134","title":"Generating Malicious Demonstration Policies to Exploit Vulnerabilities in Inverse Reinforcement Learning","year":2025,"lang":"en","type":"article","venue":"","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Exploit; Adversarial system; Reinforcement learning; Task (project management); Function (biology); Imitation","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":[],"consensus_categories":[],"category_scores_codex":[0.0005827069,0.0001541505,0.0001728403,0.0004078515,0.0002621472,0.0002149563,0.0004716499,0.00006037839,0.00002997395],"category_scores_gemma":[0.0005614391,0.0001560764,0.00003825676,0.0007683372,0.00002965031,0.0005488578,0.0004643475,0.0003090747,0.00002430156],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001956517,"about_ca_system_score_gemma":0.0001099806,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001234574,"about_ca_topic_score_gemma":0.0005225364,"domain_scores_codex":[0.9985285,0.0001778703,0.0003833839,0.00035362,0.0002199184,0.0003366853],"domain_scores_gemma":[0.999328,0.0001713639,0.00006417742,0.0003175159,0.00006116371,0.00005779493],"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":[0.000003488549,0.000007703288,0.008108714,0.00001097177,0.000004821222,0.000002907473,0.002846495,0.8873636,0.0008006725,0.09063686,0.00019761,0.01001612],"study_design_scores_gemma":[0.0002463426,0.00005921687,0.0006044446,0.00004655209,0.000002323342,0.000001609738,0.001675221,0.9940171,0.001200098,0.0008759559,0.001098413,0.00017275],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1448898,0.000009980219,0.8388644,0.002095929,0.0002312455,0.0001945536,4.596151e-8,0.0002205693,0.01349348],"genre_scores_gemma":[0.8686855,0.000002807578,0.1247597,0.001925297,0.00005108231,0.00004470112,0.000001715122,0.000006523836,0.004522646],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.7237958,"threshold_uncertainty_score":0.6364611,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0159369100862271,"score_gpt":0.2814066750714292,"score_spread":0.2654697649852021,"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."}}