{"id":"W4412617629","doi":"10.2196/71949","title":"Identifying Adverse Drug Events in Clinical Text Using Fine-Tuned Clinical Language Models: Machine Learning Study","year":2025,"lang":"en","type":"article","venue":"JMIR Formative Research","topic":"Machine Learning in Healthcare","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Linnéuniversitetet","keywords":"Preprint; Computer science; Artificial intelligence; Natural language processing; Machine learning; World Wide Web","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":true,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.004832218,0.00111451,0.0006116775,0.001919375,0.0004642878,0.00107049,0.001025381,0.00124392,0.001376186],"category_scores_gemma":[0.01604141,0.0003648451,0.001283557,0.001183253,0.0004553094,0.001865579,0.0008212837,0.002215366,0.001020055],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001208165,"about_ca_system_score_gemma":0.001047865,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008695813,"about_ca_topic_score_gemma":0.00818128,"domain_scores_codex":[0.9978975,0.0009067274,0.0002107132,0.0006928415,0.0001659858,0.0001262848],"domain_scores_gemma":[0.9791876,0.01813922,0.00063918,0.0008158592,0.0009912113,0.000226866],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001537688,0.004275355,0.1119917,0.0007484026,0.0008829016,0.0009943808,0.000875872,0.2941534,0.007652547,0.001417218,0.01037128,0.5650994],"study_design_scores_gemma":[0.00006356936,0.0003820379,0.01836205,0.00006291689,0.0001289384,0.0002945389,0.0001759488,0.9751641,0.002517636,0.001269463,0.001534877,0.00004403283],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.906133,0.003469111,0.08151776,0.001109489,0.0002509598,0.0004776809,0.002624286,0.001818315,0.002599397],"genre_scores_gemma":[0.9438522,0.000738998,0.04645805,0.0002797525,0.0001284294,0.0002098767,0.006825476,0.00008178905,0.001425531],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.008695813,"threshold_uncertainty_score":0.02555555,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2036645985138415,"score_gpt":0.5593177757802451,"score_spread":0.3556531772664036,"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."}}