{"id":"W4413215063","doi":"10.1145/3712255.3726621","title":"Discovering Blue Team Solutions for an Autonomous Cyber Operations Challenge using an Evolutionary Heuristic Search","year":2025,"lang":"en","type":"article","venue":"Proceedings of the Genetic and Evolutionary Computation Conference Companion","topic":"Advanced Software Engineering Methodologies","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Dalhousie University","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Computer science; Heuristic; Artificial intelligence","routes":{"ca_aff":true,"ca_fund":true,"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":[0.001272013,0.0007961433,0.0006391119,0.0007560711,0.0007891275,0.001015658,0.001109395,0.001429358,0.003505209],"category_scores_gemma":[0.003165798,0.0003798832,0.0005350382,0.000405097,0.0008500476,0.00090154,0.001516027,0.001006377,0.0003393705],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0008406094,"about_ca_system_score_gemma":0.001475088,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003224968,"about_ca_topic_score_gemma":0.006600496,"domain_scores_codex":[0.9995478,0.0001807496,0.0000161418,0.00008004568,0.00009014474,0.00008514927],"domain_scores_gemma":[0.9989973,0.0005764412,0.00008089076,0.00009293268,0.0001335345,0.0001190441],"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":[0.0001309471,0.0002542813,0.002888115,0.00007234608,0.00007908069,0.0002352225,0.0002665486,0.9033867,0.002813791,0.02913881,0.003236625,0.05749758],"study_design_scores_gemma":[0.0000315354,0.0001060467,0.0001891844,0.00001149081,0.00001364653,0.00002685843,0.0001267494,0.9900299,0.0004371094,0.007868997,0.001151911,0.000006457867],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3821359,0.0002709648,0.5804334,0.001257258,0.0001171946,0.0003528891,0.0001484164,0.0005693769,0.03471455],"genre_scores_gemma":[0.6734816,0.00008728058,0.3185971,0.000293237,0.00002269875,0.0003370659,0.0002089872,0.0001101044,0.006861844],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.003505209,"threshold_uncertainty_score":0.01172608,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08866842034067351,"score_gpt":0.3239043747876192,"score_spread":0.2352359544469457,"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."}}