{"id":"W4405018172","doi":"10.2196/58898","title":"Factors Associated With the Accuracy of Large Language Models in Basic Medical Science Examinations: Cross-Sectional Study","year":2024,"lang":"en","type":"article","venue":"JMIR Medical Education","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":7,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Preprint; Cross-sectional study; Medical science; Psychology; Computer science; Artificial intelligence; Medical education; Medicine; World Wide Web; 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":[],"consensus_categories":[],"category_scores_codex":[0.002879718,0.0001061955,0.000159087,0.0003267704,0.0001579577,0.00005714563,0.0002346973,0.0001515377,0.0008850893],"category_scores_gemma":[0.008151432,0.00006322237,0.0000372187,0.001471958,0.0003961933,0.0003390999,0.00003475549,0.0005317221,0.00001050588],"about_ca_system_candidate":true,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003515013,"about_ca_system_score_gemma":0.009258706,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000765997,"about_ca_topic_score_gemma":0.001095348,"domain_scores_codex":[0.9969953,0.0001160704,0.0004821154,0.0003104951,0.001844472,0.0002515245],"domain_scores_gemma":[0.9980123,0.00102234,0.00009214799,0.0002299962,0.0003862453,0.0002569667],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.00003463863,0.004073416,0.906788,0.0001363172,0.00003382674,0.00000946398,0.05385257,0.000008215056,0.00006019908,0.002641944,0.00076772,0.03159369],"study_design_scores_gemma":[0.000113374,0.0002050208,0.9625931,0.0004451882,0.00001630241,0.000007080261,0.02822338,0.00762715,0.0002661475,0.0003283779,0.0001001808,0.00007467179],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9931492,0.0003057804,0.0001315364,0.004463847,0.0007470627,0.0007850884,0.000003960459,0.00004751607,0.0003660251],"genre_scores_gemma":[0.9986652,0.00001247025,0.000004985098,0.0005460484,0.0002654773,0.0002335096,0.0000643931,0.00001220791,0.000195705],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.05580512,"threshold_uncertainty_score":0.9963579,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1567278805114249,"score_gpt":0.5171832087574989,"score_spread":0.360455328246074,"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."}}