{"id":"W4406832191","doi":"10.2196/56850","title":"Performance of ChatGPT-3.5 and ChatGPT-4 in the Taiwan National Pharmacist Licensing Examination: Comparative Evaluation Study","year":2025,"lang":"en","type":"article","venue":"JMIR Medical Education","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":6,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Pharmacist; Pharmacy; Situational ethics; Medicine; Clinical pharmacy; Medical physics; Family medicine; Medical education; Psychology; Social 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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002688884,0.00009077289,0.0001724965,0.000267484,0.0001197889,0.0000167452,0.00008741701,0.00007814342,0.000128812],"category_scores_gemma":[0.000791368,0.0000705893,0.0000197256,0.0006504545,0.0001224542,0.0001411776,0.00001543461,0.0002434569,0.000008197002],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001995037,"about_ca_system_score_gemma":0.002437869,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000109869,"about_ca_topic_score_gemma":0.00008683125,"domain_scores_codex":[0.9979606,0.0003050849,0.0004691432,0.0002038852,0.0009418328,0.0001194405],"domain_scores_gemma":[0.9986907,0.0002200949,0.0001212263,0.0001386182,0.0007608618,0.00006853153],"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":[0.0001836042,0.003835652,0.2823557,0.0003217675,0.00003220413,6.638654e-7,0.1077266,0.00001080806,0.0001519814,0.0008270521,0.002698788,0.6018551],"study_design_scores_gemma":[0.0003862061,0.0003247207,0.9036634,0.0004436211,0.00005954108,0.00001656279,0.05492903,0.03811646,0.000600404,0.0003174545,0.001061036,0.00008158836],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9839171,0.0003041425,0.00001165994,0.009260554,0.0005439864,0.001609303,5.405686e-7,0.000009662053,0.00434304],"genre_scores_gemma":[0.9977024,0.00005476647,0.00004062983,0.001368201,0.0002628178,0.0004124285,0.00004622686,0.00000366398,0.0001088418],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.6213077,"threshold_uncertainty_score":0.4324675,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1960686150432013,"score_gpt":0.5428553882530567,"score_spread":0.3467867732098553,"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."}}