{"id":"W4402937850","doi":"10.5489/cuaj.8987","title":"Beyond the hype: Unveiling the challenges of large language models in urology","year":2024,"lang":"en","type":"article","venue":"Canadian Urological Association Journal","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":1,"is_retracted":false,"has_abstract":false,"ca_institutions":"Public Health Ontario; University of Toronto","funders":"","keywords":"Urology; Medicine","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":true,"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":[0.01057884,0.0007808562,0.001662906,0.001010296,0.001332713,0.005265388,0.002823395,0.002475847,0.003434343],"category_scores_gemma":[0.07078717,0.00105208,0.001083597,0.001411579,0.003096106,0.0135917,0.004048798,0.006339687,0.001292596],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001444331,"about_ca_system_score_gemma":0.003594554,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01768435,"about_ca_topic_score_gemma":0.02870502,"domain_scores_codex":[0.9948673,0.003621301,0.0001971673,0.0005808249,0.0005466137,0.0001867922],"domain_scores_gemma":[0.9300319,0.06305595,0.001143162,0.00274498,0.002070212,0.0009537222],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.001055257,0.0003183286,0.01197634,0.0008677632,0.0005665312,0.0005858497,0.002483239,0.4117254,0.002268801,0.1719385,0.04487463,0.3513394],"study_design_scores_gemma":[0.00004256628,0.00004400019,0.0004249645,0.00005962265,0.00004460865,0.0001019968,0.0002733822,0.7821064,0.0003493174,0.2097363,0.006787285,0.00002962686],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.07859265,0.01013628,0.8312651,0.0674815,0.001213286,0.0001215177,0.002118158,0.002627454,0.006444048],"genre_scores_gemma":[0.7049236,0.004912578,0.2711603,0.007141289,0.003018826,0.0002248736,0.00195097,0.001052192,0.005615425],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01768435,"threshold_uncertainty_score":0.05594695,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07907106201243866,"score_gpt":0.3567561902958522,"score_spread":0.2776851282834135,"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."}}