{"id":"W4410976448","doi":"10.1016/j.jpurol.2025.05.030","title":"Annual updates of the European Association of Urology – European Society for Pediatric Urology (EAU-ESPU) paediatric urology guidelines: Are large-language models (LLM) better than the usual structured methodology?","year":2025,"lang":"en","type":"article","venue":"Journal of Pediatric Urology","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"McMaster University Medical Centre; London Health Sciences Centre","funders":"","keywords":"Pediatric urology; Medicine; Urology; Medical education; Pediatrics","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":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.06299056,0.0005783863,0.001379449,0.003496007,0.0009548799,0.003970987,0.003058078,0.002100665,0.004961779],"category_scores_gemma":[0.2189741,0.0006373834,0.001306906,0.003234945,0.00212979,0.004170041,0.003453863,0.003614194,0.002149587],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002550131,"about_ca_system_score_gemma":0.01188179,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.006605029,"about_ca_topic_score_gemma":0.01284076,"domain_scores_codex":[0.9477046,0.03182824,0.007003422,0.001320002,0.0116223,0.0005214356],"domain_scores_gemma":[0.7744485,0.1457331,0.01779604,0.008656862,0.04982933,0.003536248],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0002657138,0.0001172547,0.01014242,0.009508329,0.0003204081,0.0002866679,0.003588505,0.002668427,0.0008120856,0.01918053,0.3828754,0.5702344],"study_design_scores_gemma":[0.00016457,0.0004387026,0.01505467,0.03875895,0.0004296166,0.001222611,0.004368843,0.00798469,0.001893576,0.0431877,0.8862413,0.0002546837],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"commentary","genre_gemma":"empirical","genre_scores_codex":[0.05011113,0.1796885,0.2329271,0.4218586,0.01573585,0.005070785,0.01606889,0.004044157,0.07449498],"genre_scores_gemma":[0.2465797,0.1116004,0.5538984,0.04460225,0.005214889,0.005203378,0.01896025,0.002281784,0.01165895],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9370095,"threshold_uncertainty_score":0.3331299,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0888977757342952,"score_gpt":0.3895612537275523,"score_spread":0.3006634779932572,"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."}}