{"id":"W4399774223","doi":"10.1001/jamanetworkopen.2024.17641","title":"Performance of Large Language Models on Medical Oncology Examination Questions","year":2024,"lang":"en","type":"article","venue":"JAMA Network Open","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":72,"is_retracted":false,"has_abstract":true,"ca_institutions":"Institute for Clinical Evaluative Sciences; Vector Institute; University of Toronto; Princess Margaret Cancer Centre; University Health Network","funders":"","keywords":"Clinical Oncology; Medicine; Harm; Oncology; Internal medicine; Clinical trial; Set (abstract data type); Family medicine; Psychology; Cancer; Computer science","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001729925,0.0000652891,0.000167789,0.00006420047,0.00007115011,0.00002855504,0.000143948,0.0001775232,0.0006323105],"category_scores_gemma":[0.0001934733,0.00005394039,0.000030153,0.0002756316,0.00003507053,0.0002051044,0.00005953513,0.0002960924,0.000132494],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001213376,"about_ca_system_score_gemma":0.000494229,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0003348463,"about_ca_topic_score_gemma":0.0001590675,"domain_scores_codex":[0.9989836,0.0001213658,0.0003020133,0.000167111,0.0002357861,0.0001901741],"domain_scores_gemma":[0.9992532,0.0003206894,0.00004534739,0.0001779758,0.00009266104,0.000110112],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0002238953,0.0002997255,0.003667801,0.0002401536,0.00003692761,0.00004666705,0.006393342,0.001406936,0.0000360811,0.02699409,0.02285287,0.9378015],"study_design_scores_gemma":[0.0001883184,0.001899257,0.0222547,0.003828553,0.00007650085,0.00007182288,0.002338739,0.926406,0.001230558,0.002218294,0.03931423,0.0001729968],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9368344,0.0007782402,0.0007961034,0.01023311,0.0008245359,0.0004611671,0.000004382882,0.00004720401,0.05002084],"genre_scores_gemma":[0.9948708,0.0005548255,0.0003890495,0.001314038,0.001111251,0.00005945489,0.00006413647,0.00001236824,0.001624013],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.9376285,"threshold_uncertainty_score":0.6923357,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1192965486447497,"score_gpt":0.4667730081895255,"score_spread":0.3474764595447757,"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."}}