{"id":"W7125517580","doi":"10.1109/cars67163.2025.11337724","title":"Autonomous Multi-agent Cyber Defense: A Novel Approach Using Reinforcement Learning with Hierarchical LLM Critics","year":2025,"lang":"en","type":"article","venue":"","topic":"Military Defense Systems Analysis","field":"Engineering","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Defence Research and Development Canada; National Research Council Canada","funders":"","keywords":"Reinforcement learning; Action (physics); Control (management); Key (lock); Matching (statistics)","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"],"consensus_categories":[],"category_scores_codex":[0.0002924411,0.0002945466,0.0003928188,0.0002935514,0.0001582894,0.00007663679,0.0001513368,0.0001165244,0.00005667992],"category_scores_gemma":[0.00004508168,0.0002551552,0.0001306692,0.000484846,0.00005572015,0.0001363925,0.00007685397,0.0003982921,0.00001643573],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003122673,"about_ca_system_score_gemma":0.00006711342,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000461919,"about_ca_topic_score_gemma":0.00007925967,"domain_scores_codex":[0.9983961,0.00004771681,0.0004822435,0.0003536853,0.0002654774,0.000454806],"domain_scores_gemma":[0.9993126,0.00006212647,0.0000294377,0.0003994654,0.00008339698,0.0001129535],"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":[0.000007285427,0.00006444897,0.0005345645,0.0002140915,0.0003687949,0.00000935426,0.0004237804,0.9920647,0.001403581,0.004589696,0.00009399901,0.0002257242],"study_design_scores_gemma":[0.0006167635,0.00002711252,0.0001889104,0.00007066168,0.0001470499,0.00002650603,0.0006012536,0.9930724,0.0002030877,0.000005440962,0.004738353,0.0003024881],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.05395854,0.000247035,0.9195634,0.00001829015,0.0001007894,0.000275533,0.000002465837,0.0004124593,0.02542154],"genre_scores_gemma":[0.869236,0.000006028137,0.1237182,0.00009299331,0.00003939242,0.00003073142,0.00001714635,0.00004557147,0.006813924],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.8152775,"threshold_uncertainty_score":0.99999,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01943326890434323,"score_gpt":0.2374218119093884,"score_spread":0.2179885430050452,"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."}}