{"id":"W4408212176","doi":"10.1055/a-2551-2131","title":"Solving Complex Pediatric Surgical Case Studies: A Comparative Analysis of Copilot, ChatGPT-4, and Experienced Pediatric Surgeons' Performance","year":2025,"lang":"en","type":"article","venue":"European Journal of Pediatric Surgery","topic":"Cardiac, Anesthesia and Surgical Outcomes","field":"Medicine","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"Hospital for Sick Children","funders":"","keywords":"Medicine; General surgery; Pediatric Surgeon; Pediatric surgery; Surgery","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":[0.006013509,0.0004520986,0.0004269549,0.001517397,0.0002818193,0.001343977,0.0009277756,0.00056576,0.001934979],"category_scores_gemma":[0.043595,0.0002448226,0.0005409425,0.001002125,0.0006156523,0.001273302,0.001505165,0.0006337545,0.0004279795],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009277288,"about_ca_system_score_gemma":0.0009599559,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002749478,"about_ca_topic_score_gemma":0.003476268,"domain_scores_codex":[0.9956619,0.002212591,0.00050223,0.0004788634,0.0009235168,0.000220949],"domain_scores_gemma":[0.9523435,0.03325579,0.006275625,0.001813737,0.003681224,0.002630136],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.001723841,0.001386772,0.8257326,0.0006032239,0.000352991,0.001075291,0.01081611,0.01949385,0.003104978,0.0004789656,0.002758405,0.1324729],"study_design_scores_gemma":[0.0001898834,0.006374507,0.7786412,0.0004807446,0.0003145291,0.003599498,0.02146647,0.1706122,0.009384893,0.001012408,0.007725908,0.0001977433],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9974483,0.0001313991,0.00122743,0.00008738763,0.000006703447,0.00004849602,0.0001192239,0.00006358852,0.0008674968],"genre_scores_gemma":[0.9941507,0.0002028185,0.00486267,0.00004126686,0.000009315229,0.00004116226,0.000394391,0.00002005203,0.0002775308],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.9939865,"threshold_uncertainty_score":0.03180289,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06479984516485819,"score_gpt":0.3323002339807762,"score_spread":0.267500388815918,"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."}}