{"id":"W4415706004","doi":"10.36227/techrxiv.176184397.77169395/v1","title":"Security Knowledge Dilution in Large Language Models: How Irrelevant Context Degrades Critical Domain Expertise","year":2025,"lang":"","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Vector Institute","funders":"","keywords":"Context (archaeology); Phenomenon; Domain (mathematical analysis); Domain knowledge; Knowledge acquisition; Term (time); Relevance (law)","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":[0.00377056,0.0007161071,0.0005978788,0.0004047249,0.0005930634,0.001545411,0.001162144,0.001088068,0.001560206],"category_scores_gemma":[0.05952189,0.0005972041,0.000372118,0.0002580133,0.001289598,0.00353123,0.002609944,0.001812748,0.0004241475],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006554552,"about_ca_system_score_gemma":0.0006725152,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001924718,"about_ca_topic_score_gemma":0.002093654,"domain_scores_codex":[0.9964593,0.001943485,0.0002347951,0.0006352951,0.0005125906,0.0002144985],"domain_scores_gemma":[0.9627305,0.03062003,0.001448691,0.003540554,0.000809064,0.000851061],"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":[0.005566868,0.001954044,0.07736301,0.0009916914,0.0005133165,0.001750097,0.01234948,0.3315929,0.3116929,0.006456265,0.004150523,0.2456191],"study_design_scores_gemma":[0.0002333376,0.002369455,0.02237488,0.00009509411,0.0002691884,0.0007707021,0.00180643,0.7972517,0.1518529,0.01772071,0.00508604,0.0001696429],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9705396,0.000185068,0.02674598,0.0003348663,0.00002446221,0.00005403112,0.00008485054,0.0007644721,0.001266777],"genre_scores_gemma":[0.9917521,0.00005014765,0.007554102,0.0001143405,0.00001287472,0.00003271093,0.0001196274,0.00009568177,0.0002685295],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.00377056,"threshold_uncertainty_score":0.01994085,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02692883516655278,"score_gpt":0.3072965362562432,"score_spread":0.2803677010896904,"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."}}