{"id":"W4390919625","doi":"10.1148/radiol.232411","title":"Comparing GPT-3.5 and GPT-4 Accuracy and Drift in <i>Radiology</i> Diagnosis Please Cases","year":2024,"lang":"en","type":"article","venue":"Radiology","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":52,"is_retracted":false,"has_abstract":false,"ca_institutions":"London Health Sciences Centre; Western University","funders":"","keywords":"Medicine; Medical physics; Radiology; Nuclear medicine","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.006817091,0.0007582682,0.0007022105,0.003315636,0.0005431392,0.00214697,0.00112897,0.002416323,0.001216829],"category_scores_gemma":[0.04465181,0.0003238663,0.001062678,0.00142163,0.0007506619,0.00157711,0.001364016,0.0009773858,0.001253007],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001285162,"about_ca_system_score_gemma":0.001308764,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01501707,"about_ca_topic_score_gemma":0.008883803,"domain_scores_codex":[0.9965497,0.0007372625,0.0004279116,0.0008644139,0.001067415,0.0003533036],"domain_scores_gemma":[0.9836972,0.008051809,0.001489898,0.001566876,0.004569532,0.0006247409],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.00397601,0.0002341229,0.6841055,0.0003604322,0.001029932,0.0007139496,0.0005943194,0.06124625,0.008809415,0.001411068,0.01704155,0.2204774],"study_design_scores_gemma":[0.0003161831,0.0017536,0.454289,0.000254456,0.001194677,0.005363168,0.001141117,0.4801561,0.03262618,0.008065324,0.01456571,0.0002744094],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.958594,0.002622945,0.02251574,0.001355065,0.0005337406,0.0001046954,0.006085179,0.003435662,0.004752979],"genre_scores_gemma":[0.9785656,0.0003378777,0.01384298,0.0002109414,0.0001385571,0.00003203338,0.005444828,0.000294359,0.001132748],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01501707,"threshold_uncertainty_score":0.03605264,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1689424278669887,"score_gpt":0.4260712340969973,"score_spread":0.2571288062300087,"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."}}