{"id":"W7127150443","doi":"10.1109/trustcom66490.2025.00256","title":"L2M-AID: Autonomous Cyber-Physical Defense by Fusing Semantic Reasoning of Large Language Models with Multi-Agent Reinforcement Learning","year":2025,"lang":"en","type":"article","venue":"","topic":"Smart Grid Security and Resilience","field":"Engineering","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"Thales (Canada)","funders":"","keywords":"Reinforcement learning; Leverage (statistics); Process (computing); Adversary; Matching (statistics); State (computer science); Key (lock); Function (biology)","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":[0.000130817,0.0001850412,0.0002541892,0.00009178552,0.0001157652,0.00002922054,0.0001279679,0.00006379998,0.00002046896],"category_scores_gemma":[0.00001912837,0.0001556096,0.00006466045,0.0002315515,0.00004003942,0.0001482518,0.00007130914,0.0002608457,0.000008823152],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00007081727,"about_ca_system_score_gemma":0.00003037408,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000165803,"about_ca_topic_score_gemma":0.00004871693,"domain_scores_codex":[0.9989989,0.0000252312,0.000221261,0.0002066407,0.0001874593,0.0003605153],"domain_scores_gemma":[0.9995989,0.00005249407,0.00003998188,0.0002118004,0.0000321282,0.00006472658],"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":[0.00001615278,0.00006970885,0.0005523721,0.0001719864,0.00007706346,0.00001527456,0.004775598,0.9784877,0.01215846,0.003037882,0.0002448765,0.0003929534],"study_design_scores_gemma":[0.0006500396,0.00004279818,0.0001405769,0.0002158911,0.00003793198,0.000003839208,0.001848477,0.9698883,0.0265898,0.000009842981,0.000378697,0.0001938293],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7123919,0.0003797656,0.279339,0.00001464628,0.00008456253,0.0001536409,0.000002458762,0.000247028,0.007387024],"genre_scores_gemma":[0.9965192,0.00002783077,0.002041746,0.00004201945,0.00002518817,0.00001077204,0.0000141339,0.00002230364,0.001296816],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.2841274,"threshold_uncertainty_score":0.6345575,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.005626917600127772,"score_gpt":0.2268651427682939,"score_spread":0.2212382251681662,"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."}}