{"id":"W4410876142","doi":"10.1007/s10207-025-01066-4","title":"A Data-centric approach for safe and secure large language models against threatening and toxic content","year":2025,"lang":"en","type":"article","venue":"International Journal of Information Security","topic":"Hate Speech and Cyberbullying Detection","field":"Computer Science","cited_by":3,"is_retracted":false,"has_abstract":false,"ca_institutions":"Université du Québec à Chicoutimi","funders":"","keywords":"Computer science; Cryptography; Computer security; Internet privacy","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":[0.0006503217,0.00007987649,0.0001217889,0.0002844502,0.00007927926,0.0003263124,0.0005212439,0.00004973054,0.000001217858],"category_scores_gemma":[0.0001172608,0.00006767852,0.00003635996,0.0001166017,0.00001559465,0.003284036,0.0002484657,0.0001389027,6.299811e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00006100207,"about_ca_system_score_gemma":0.00006553856,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000007581378,"about_ca_topic_score_gemma":0.000003424847,"domain_scores_codex":[0.9991101,0.00002349466,0.0003894735,0.00009613139,0.0002698276,0.0001109909],"domain_scores_gemma":[0.9990396,0.00005121658,0.0002814775,0.0001375451,0.0004351902,0.00005492387],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0005352306,0.0003148319,0.0009368197,0.0002950072,0.0007825797,0.00002903887,0.02143326,0.001810759,0.0006524885,0.6310655,0.004179295,0.3379652],"study_design_scores_gemma":[0.002266628,0.00006122565,0.0003709709,0.00008782477,0.0000177904,0.0001499783,0.001476222,0.974417,0.0007007282,0.005319181,0.01500811,0.0001243199],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.06102405,0.000482263,0.9354936,0.0006338308,0.0003563643,0.0001441581,0.00008723266,0.00001939479,0.001759134],"genre_scores_gemma":[0.9711754,0.0002400441,0.02775321,0.000678547,0.00006200976,0.000003589442,0.00006553019,0.0000019553,0.00001975609],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9726062,"threshold_uncertainty_score":0.3146638,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02298082640046019,"score_gpt":0.2729595508045289,"score_spread":0.2499787244040688,"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."}}