{"id":"W4414413212","doi":"10.21203/rs.3.rs-7511791/v1","title":"Fine-grained Insider Threat Detection with Large Language Models: A Comparative Study","year":2025,"lang":"en","type":"preprint","venue":"Research Square","topic":"Information and Cyber Security","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"Polytechnique Montréal","funders":"Compute Canada","keywords":"Insider threat; Insider; Generative grammar; Generative model; Behavioral analysis; Stability (learning theory); Behavioral economics","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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001536062,0.000276826,0.000380156,0.0006574348,0.0004479453,0.0005723179,0.001180294,0.0001809018,0.00002944268],"category_scores_gemma":[0.00006031747,0.0002194269,0.00009500365,0.001043263,0.00006198823,0.0006124343,0.002452987,0.001621509,0.00006333096],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002458655,"about_ca_system_score_gemma":0.0005661673,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0007517803,"about_ca_topic_score_gemma":0.007549042,"domain_scores_codex":[0.9963685,0.0007189411,0.0003239374,0.0006365529,0.001377902,0.0005741977],"domain_scores_gemma":[0.9972213,0.000196722,0.000105002,0.001315761,0.001026388,0.0001348077],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0005841656,0.003718545,0.00219554,0.002006012,0.0007880967,0.0003754856,0.8188534,0.01400016,0.00005441177,0.1327845,0.004276739,0.02036287],"study_design_scores_gemma":[0.003145691,0.001230149,0.005955647,0.0007357869,0.00002571057,0.00001145799,0.0377461,0.9351285,0.001610769,0.01229092,0.001286942,0.0008323862],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4053048,0.0005952904,0.531413,0.0006814887,0.0003859055,0.005906845,0.0001830092,0.0008119554,0.05471779],"genre_scores_gemma":[0.9973671,0.000009294325,0.001062624,0.00005433359,0.00005494007,0.0004560728,0.0000454522,0.000008006108,0.0009421795],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9211283,"threshold_uncertainty_score":0.8947971,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0821520498102914,"score_gpt":0.3995551456815371,"score_spread":0.3174030958712457,"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."}}