{"id":"W4410477368","doi":"10.1542/pedsos.2025-000485","title":"PedMedQA: Comparing Large Language Model Accuracy in Pediatric and Adult Medicine","year":2025,"lang":"en","type":"article","venue":"Pediatrics Open Science","topic":"Machine Learning in Healthcare","field":"Computer Science","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"Mila - Quebec Artificial Intelligence Institute; Montreal Children's Hospital; McGill University; Polytechnique Montréal; Centre for Advancing Health Outcomes; McGill University Health Centre","funders":"Fonds de Recherche du Québec - Santé; Canadian Institutes of Health Research","keywords":"Computer science; Natural language processing","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.01090045,0.001078538,0.0008989221,0.00205262,0.0004453254,0.002032192,0.002075288,0.0015401,0.01035983],"category_scores_gemma":[0.07332753,0.0004258657,0.001218901,0.001365401,0.0006470082,0.002163364,0.002092943,0.002288786,0.003251772],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001707075,"about_ca_system_score_gemma":0.002919404,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01212918,"about_ca_topic_score_gemma":0.01055171,"domain_scores_codex":[0.994055,0.003536268,0.0004980542,0.001027753,0.0007203031,0.0001627512],"domain_scores_gemma":[0.9585056,0.03395581,0.001303041,0.002075447,0.002985644,0.00117444],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.00703629,0.0009956597,0.1508739,0.003795657,0.002713811,0.0006648509,0.0006941747,0.151684,0.00212475,0.00578767,0.336256,0.3373733],"study_design_scores_gemma":[0.003134726,0.003311219,0.06066541,0.001629504,0.0009943672,0.002383758,0.0009452584,0.7929005,0.008334302,0.02478153,0.1005468,0.0003726255],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5447612,0.02999627,0.1580339,0.01754089,0.003201858,0.001304532,0.1899111,0.0331934,0.02205691],"genre_scores_gemma":[0.7915749,0.002834147,0.07616238,0.00303833,0.000665454,0.0006569038,0.1209911,0.001349455,0.002727331],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01212918,"threshold_uncertainty_score":0.05764776,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02644698040996564,"score_gpt":0.3821723891951043,"score_spread":0.3557254087851387,"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."}}