{"id":"W4405623292","doi":"10.2196/63731","title":"Qwen-2.5 Outperforms Other Large Language Models in the Chinese National Nursing Licensing Examination: Retrospective Cross-Sectional Comparative Study","year":2024,"lang":"en","type":"article","venue":"JMIR Medical Informatics","topic":"Nursing Diagnosis and Documentation","field":"Nursing","cited_by":26,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Preprint; Cross-sectional study; Medicine; Nursing; Computer science; World Wide Web","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":[0.002025766,0.0002134475,0.0002362779,0.0002813193,0.0002894482,0.000583234,0.0002393705,0.0001502093,0.0002336198],"category_scores_gemma":[0.0001354399,0.0001412198,0.00008123176,0.0007706659,0.0001553731,0.001434381,0.00002855225,0.0006636086,0.00007795695],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000622735,"about_ca_system_score_gemma":0.00007986028,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002627752,"about_ca_topic_score_gemma":0.0000731566,"domain_scores_codex":[0.9966577,0.0001677116,0.0007243167,0.0001931045,0.001938328,0.0003187866],"domain_scores_gemma":[0.9990663,0.0003889457,0.0001305501,0.000164372,0.0001649632,0.00008490513],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"observational","study_design_scores_codex":[0.0001155341,0.00205004,0.1429702,0.0001232073,0.0000694659,0.0000194537,0.8357128,0.0006819142,0.00001145907,0.007092209,0.003865205,0.007288534],"study_design_scores_gemma":[0.001354036,0.0002721364,0.6587707,0.0003966209,0.00001500898,0.0000432106,0.05118733,0.2846183,0.0000199981,0.002883341,0.0002203706,0.0002189242],"study_design_candidate":"qualitative","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9793842,0.000116796,0.000919909,0.0005057136,0.001213151,0.0008265876,0.00002958648,0.0001266014,0.01687739],"genre_scores_gemma":[0.9979711,0.000001729759,0.000201367,0.001017865,0.0005370777,0.000104779,0.00008108569,0.00001904913,0.00006600627],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.7845255,"threshold_uncertainty_score":0.5758777,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02972743349575719,"score_gpt":0.4125477591948185,"score_spread":0.3828203256990613,"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."}}