{"id":"W4389577763","doi":"10.2196/48514","title":"Performance of ChatGPT on the Chinese Postgraduate Examination for Clinical Medicine: Survey Study","year":2023,"lang":"en","type":"article","venue":"JMIR Medical Education","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":36,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Medical education; Medicine; Family medicine; Traditional Chinese medicine; Traditional medicine; Psychology; Alternative medicine; Pathology","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":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.006905428,0.000126177,0.0002950632,0.0001997695,0.0001427158,0.000006044543,0.0001717576,0.0001458665,0.0002278288],"category_scores_gemma":[0.0155996,0.00007398539,0.00006440211,0.0009827566,0.0001932832,0.00005872,0.00001966681,0.0002916291,0.00009786648],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005909575,"about_ca_system_score_gemma":0.001094347,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0003772593,"about_ca_topic_score_gemma":0.00009625432,"domain_scores_codex":[0.9975349,0.0004050472,0.0008342552,0.0002784631,0.000740436,0.0002069229],"domain_scores_gemma":[0.9958759,0.002628265,0.0002221944,0.0003949679,0.0006636424,0.0002150252],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0002636393,0.001478974,0.6614143,0.0001332311,0.00002176281,3.577823e-7,0.004967968,5.480284e-7,0.0000129987,0.0001065571,0.02304062,0.3085591],"study_design_scores_gemma":[0.0001737004,0.00243256,0.9888904,0.0002385098,0.00002416026,0.000002730949,0.00423945,0.003110108,0.00006122032,0.0001577864,0.0006039144,0.0000654342],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9676748,0.00003564432,0.000009937263,0.0274765,0.002233078,0.002107436,0.000002987181,0.00005100689,0.0004086032],"genre_scores_gemma":[0.9950815,0.0002917912,0.00001008042,0.00162419,0.001342193,0.0006696746,0.0002558016,0.0000186352,0.0007060858],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3274762,"threshold_uncertainty_score":0.9926924,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2771935890227568,"score_gpt":0.5664957622394303,"score_spread":0.2893021732166735,"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."}}