{"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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow","research_integrity"],"consensus_categories":[],"category_scores_codex":[0.02045525,0.0002701218,0.0006440093,0.001164837,0.0006449078,0.0001340576,0.001974784,0.0001893896,0.00003198037],"category_scores_gemma":[0.001934473,0.0002485744,0.000221299,0.002578225,0.0001619745,0.001395522,0.002825223,0.004942969,0.0001340997],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003607427,"about_ca_system_score_gemma":0.000532997,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001836005,"about_ca_topic_score_gemma":0.0009351447,"domain_scores_codex":[0.9853323,0.009528238,0.001825102,0.0009174963,0.001379675,0.001017246],"domain_scores_gemma":[0.9950511,0.002954063,0.0003109685,0.001015145,0.0004197164,0.0002490086],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0001667419,0.001459337,0.8864262,0.0002848208,0.00008145229,0.0001353517,0.04038687,0.005433561,0.000008102837,0.002240747,0.0002691608,0.06310768],"study_design_scores_gemma":[0.001903694,0.0003263136,0.2800121,0.0003295391,0.000004075309,0.000003460119,0.004771767,0.7109453,0.00000621983,0.001366311,0.0001403362,0.0001908645],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9674891,0.0002815505,0.02771978,0.0006887711,0.0006148859,0.001515907,0.000003309203,0.0001806558,0.001506097],"genre_scores_gemma":[0.9955744,0.00002846318,0.003043289,0.00007805855,0.00009229497,0.0001117931,0.000008992665,0.00002375465,0.00103898],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.7055117,"threshold_uncertainty_score":0.9999967,"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."}}