{"id":"W4409283616","doi":"10.1038/s41586-025-08869-4","title":"Towards accurate differential diagnosis with large language models","year":2025,"lang":"en","type":"article","venue":"Nature","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":173,"is_retracted":false,"has_abstract":true,"ca_institutions":"Google (Canada)","funders":"","keywords":"McNemar's test; Differential diagnosis; Test (biology); Medicine; Process (computing); Computer science; Medical physics; Differential (mechanical device); Pathology; Statistics; Mathematics","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":[0.00006615582,0.00008944419,0.0001405456,0.00008490888,0.00007731827,0.00001924615,0.00006517832,0.0005282758,0.0002314996],"category_scores_gemma":[0.00009173783,0.00006241497,0.00004024285,0.0002504849,0.00001878687,0.00007359181,0.00001911065,0.001022983,0.00001716011],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005689001,"about_ca_system_score_gemma":0.0002017472,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002057249,"about_ca_topic_score_gemma":0.0002029281,"domain_scores_codex":[0.9993325,0.00001823161,0.0001283797,0.0001729234,0.0001545455,0.0001934054],"domain_scores_gemma":[0.9995205,0.0000470352,0.0000284643,0.0002005332,0.0001376943,0.00006577775],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.00209323,0.002128656,0.1370639,0.001449686,0.0004901431,0.000206542,0.02149621,0.0001283364,0.001844889,0.07549734,0.1516274,0.6059737],"study_design_scores_gemma":[0.002336571,0.00191961,0.1950663,0.004457484,0.001540083,0.00009387603,0.03499611,0.03166713,0.5454227,0.03768719,0.14304,0.001773001],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.973137,0.00193385,0.003147483,0.01344519,0.0006672608,0.0003254953,0.00001717784,0.00007953468,0.007247048],"genre_scores_gemma":[0.993676,0.0001400692,0.0002269917,0.003952593,0.0002662024,0.00005663683,0.00005674126,0.00000895169,0.001615804],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.6042007,"threshold_uncertainty_score":0.4444407,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05125932754292,"score_gpt":0.4184976255190135,"score_spread":0.3672382979760935,"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."}}