{"id":"W4399851271","doi":"10.2196/52784","title":"Influence of Model Evolution and System Roles on ChatGPT’s Performance in Chinese Medical Licensing Exams: Comparative Study","year":2024,"lang":"en","type":"article","venue":"JMIR Medical Education","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":19,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Psychology","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":[0.01976573,0.0008626785,0.0006896773,0.001705271,0.0006720915,0.002193562,0.001050844,0.0009550474,0.001132856],"category_scores_gemma":[0.09808514,0.0004286211,0.001262763,0.000976676,0.001114714,0.003342332,0.001571711,0.0009421352,0.0003031323],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002361909,"about_ca_system_score_gemma":0.001478445,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01373452,"about_ca_topic_score_gemma":0.007330849,"domain_scores_codex":[0.9900629,0.006051269,0.0006747564,0.001242861,0.001464229,0.0005040371],"domain_scores_gemma":[0.9080516,0.07122704,0.00617685,0.005563298,0.006694817,0.002286338],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.001864347,0.0007984637,0.9061662,0.0002400401,0.00042444,0.0003920207,0.007633129,0.01533905,0.001582942,0.0004645126,0.0004871164,0.06460772],"study_design_scores_gemma":[0.0001090987,0.003092338,0.7102894,0.0001369506,0.0009014616,0.0005450937,0.006496178,0.2724299,0.003687754,0.0008149854,0.001352434,0.0001444254],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9979971,0.0001045457,0.001254071,0.00004347237,0.000006968719,0.00003939766,0.0000328209,0.00003734947,0.0004841708],"genre_scores_gemma":[0.9988784,0.00004462491,0.000777853,0.0000102358,0.000004959217,0.0000221761,0.00009807594,0.00001049005,0.0001531269],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01976573,"threshold_uncertainty_score":0.1045324,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07091824250082696,"score_gpt":0.4584233130009801,"score_spread":0.3875050705001531,"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."}}