{"id":"W4404902547","doi":"10.1016/j.clinimag.2024.110382","title":"Comparative diagnostic accuracy of GPT-4o and LLaMA 3-70b: Proprietary vs. open-source large language models in radiology","year":2024,"lang":"en","type":"article","venue":"Clinical Imaging","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":19,"is_retracted":false,"has_abstract":false,"ca_institutions":"McGill University; McGill University Health Centre; London Health Sciences Centre","funders":"","keywords":"Medicine; Radiology; Nuclear medicine; Medical physics","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":[0.001141617,0.0001032955,0.0004649223,0.0000929162,0.00004061171,0.00003623862,0.0001182416,0.00008056366,0.00006030625],"category_scores_gemma":[0.002513419,0.00008599349,0.00005457587,0.0001930157,0.0002093215,0.0002387567,0.0001235753,0.0004566733,0.00003115189],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003404434,"about_ca_system_score_gemma":0.000235862,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0007893126,"about_ca_topic_score_gemma":0.00009807948,"domain_scores_codex":[0.9983974,0.0001945879,0.0007468292,0.0003389183,0.00008809826,0.0002341771],"domain_scores_gemma":[0.9948208,0.004680818,0.00008574356,0.0002101045,0.00007996782,0.000122615],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000391156,0.0006624397,0.5333769,0.0006318231,0.00007361075,0.000281007,0.02859013,0.0001144056,0.0003810777,0.001879485,0.004879773,0.4287382],"study_design_scores_gemma":[0.001552371,0.001407464,0.2205684,0.00522109,0.0003675493,0.0004385637,0.037968,0.6868775,0.004996346,0.02381782,0.0159971,0.0007877787],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9808619,0.007922959,0.002012142,0.007293039,0.0004506574,0.0006144575,0.000006476289,0.00004045074,0.0007979393],"genre_scores_gemma":[0.9967486,0.0007537942,0.0005631248,0.001431862,0.0002946699,0.00003007109,0.00002830548,0.00001439474,0.0001351282],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.6867632,"threshold_uncertainty_score":0.3506713,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2348965275187137,"score_gpt":0.5358115507626536,"score_spread":0.3009150232439399,"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."}}