{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001312021,0.0009326737,0.000928063,0.0004932013,0.0004437126,0.0008853345,0.002102643,0.001217729,0.002512393],"category_scores_gemma":[0.003327565,0.0005738575,0.0005611387,0.0002929014,0.00112431,0.001084284,0.001612696,0.001511691,0.0004056633],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001039175,"about_ca_system_score_gemma":0.001211378,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005720936,"about_ca_topic_score_gemma":0.005928337,"domain_scores_codex":[0.999543,0.0001683819,0.00002296299,0.0001050481,0.0001003836,0.00006036514],"domain_scores_gemma":[0.998521,0.0008011416,0.0002081969,0.0001410414,0.0002106533,0.0001179167],"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":[0.00005510196,0.00005455357,0.0008701626,0.00004074349,0.0000528337,0.0001156255,0.0001018897,0.9591396,0.002057109,0.01225423,0.0007291179,0.024529],"study_design_scores_gemma":[0.000004728657,0.0000107462,0.00002358296,0.000001615776,0.000003063185,0.000004840102,0.00000243981,0.9982717,0.0001291777,0.001417885,0.0001280088,0.000002226737],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02407776,0.0001757093,0.9712321,0.0004553017,0.0000484573,0.00005804869,0.00002800628,0.0006991975,0.003225503],"genre_scores_gemma":[0.8845214,0.00008780547,0.1106841,0.0002571597,0.00004883151,0.0001326424,0.00005466632,0.0000897748,0.004123704],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.005720936,"threshold_uncertainty_score":0.01137531,"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."}}