{"id":"W4416222729","doi":"10.2196/73469","title":"Evaluating the Performance of DeepSeek-R1 and DeepSeek-V3 Versus OpenAI Models in the Chinese National Medical Licensing Examination: Cross-Sectional Comparative Study","year":2025,"lang":"en","type":"article","venue":"JMIR Medical Education","topic":"Academic integrity and plagiarism","field":"Social Sciences","cited_by":8,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Licensure","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":[0.02197122,0.0008064144,0.0008785361,0.001339957,0.0005573268,0.001332336,0.001235813,0.0008654881,0.002585765],"category_scores_gemma":[0.05073969,0.0004006279,0.001758882,0.0008443096,0.0007376675,0.002653499,0.001821812,0.001727534,0.0009029119],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001603852,"about_ca_system_score_gemma":0.001989684,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01300738,"about_ca_topic_score_gemma":0.01172344,"domain_scores_codex":[0.9932519,0.003414147,0.0006133207,0.001140356,0.001199981,0.0003802345],"domain_scores_gemma":[0.959047,0.02480405,0.004150846,0.003004824,0.007034701,0.001958572],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0007614978,0.0008253212,0.961452,0.0001801962,0.0004191866,0.00008853981,0.001416665,0.002983559,0.0003039497,0.0002668319,0.001922965,0.02937939],"study_design_scores_gemma":[0.0001815837,0.002911519,0.9008674,0.0002793575,0.0009346911,0.0003278562,0.003690077,0.08300575,0.002649439,0.00106801,0.003942303,0.0001419819],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9963551,0.0002805904,0.001387014,0.0002289164,0.00002545292,0.00009777631,0.0008506567,0.00005263175,0.0007218676],"genre_scores_gemma":[0.9958007,0.0001381483,0.001391542,0.000110695,0.0000219947,0.0001529174,0.001862025,0.00002252256,0.0004994977],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.02197122,"threshold_uncertainty_score":0.1161963,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.09535028815819191,"score_gpt":0.5095570624048554,"score_spread":0.4142067742466635,"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."}}