{"id":"W4414368365","doi":"10.2196/68070","title":"Evaluating the Potential and Accuracy of ChatGPT-3.5 and 4.0 in Medical Licensing and In-Training Examinations: Systematic Review and Meta-Analysis","year":2025,"lang":"en","type":"article","venue":"JMIR Medical Education","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":15,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Reliability (semiconductor); Set (abstract data type); Focus (optics); MEDLINE; Component (thermodynamics); Medical diagnosis","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":[0.03578936,0.002390874,0.01572159,0.007978067,0.0008015186,0.00424844,0.00288501,0.002975542,0.003339525],"category_scores_gemma":[0.1185483,0.001595361,0.04187112,0.008444401,0.001378598,0.002926625,0.00234021,0.002114598,0.000338164],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.003207227,"about_ca_system_score_gemma":0.005772113,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008317269,"about_ca_topic_score_gemma":0.01875243,"domain_scores_codex":[0.9702091,0.01182294,0.01110955,0.002770599,0.003524338,0.0005635696],"domain_scores_gemma":[0.8770131,0.1007569,0.01450213,0.002521444,0.004704284,0.0005021656],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"systematic_review","study_design_gemma":"meta_analysis","study_design_scores_codex":[0.001351861,0.00003670972,0.007880769,0.6237233,0.3443817,0.00007531089,0.00018053,0.0004107906,0.0001298235,0.0002001285,0.0007167008,0.02091232],"study_design_scores_gemma":[0.000484651,0.0002558736,0.006600579,0.1043771,0.884841,0.0001180935,0.0001072611,0.0002035661,0.0001751753,0.0002964127,0.002498919,0.00004137779],"study_design_candidate":"meta_analysis","study_design_consensus":null,"genre_codex":"review","genre_gemma":"empirical","genre_scores_codex":[0.003231183,0.9946654,0.0005748443,0.0002108389,0.0001225178,0.0003520195,0.0005839845,0.00001877408,0.0002405072],"genre_scores_gemma":[0.1746956,0.8145957,0.004328084,0.001627394,0.0003887749,0.002496449,0.001493098,0.00005430756,0.0003205445],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.03578936,"threshold_uncertainty_score":0.1892745,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2479325627637239,"score_gpt":0.5553943139980585,"score_spread":0.3074617512343346,"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."}}