{"id":"W4417049848","doi":"10.1177/22925503251400370","title":"The Use of Large Language Models in Postgraduate Plastic Surgery Training: A National Survey of Plastic Surgery Residents","year":2025,"lang":"en","type":"article","venue":"Plastic Surgery","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto; University of Ottawa","funders":"","keywords":"Plastic surgery; Leverage (statistics); Descriptive statistics; MEDLINE; Interview; English language","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":true,"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":[0.00153039,0.0001435973,0.000157464,0.0007470655,0.0009305173,0.0006226472,0.0006485077,0.0003503441,0.002083502],"category_scores_gemma":[0.006574915,0.0002235201,0.0001790543,0.001256258,0.0007316498,0.0005282155,0.000717448,0.0004954082,0.0002632634],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00303774,"about_ca_system_score_gemma":0.00678629,"about_ca_topic_candidate":true,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.5123454,"about_ca_topic_score_gemma":0.6519297,"domain_scores_codex":[0.9985165,0.0002278449,0.0001075741,0.0001073751,0.0007571594,0.0002835324],"domain_scores_gemma":[0.9950625,0.0008108286,0.001737997,0.00007074139,0.001073287,0.001244589],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.000048143,0.0001113576,0.9706003,0.0001915951,0.00001404428,0.0001795126,0.01404263,0.00005933284,0.0006581406,0.00002855955,0.001624417,0.01244204],"study_design_scores_gemma":[0.000002916259,0.000107619,0.9780653,0.00008307129,0.000007288585,0.0002252172,0.01963233,0.0001326488,0.00009957534,0.000008237457,0.001622599,0.00001314692],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9982182,0.0001790134,0.00005778141,0.0002999083,0.000005772625,0.00002682054,0.000276383,0.000006174734,0.0009298907],"genre_scores_gemma":[0.9984654,0.000441086,0.0002322046,0.0002219912,0.000005363498,0.00002657043,0.0002095438,0.000003283538,0.0003945816],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.9984696,"threshold_uncertainty_score":0.9810531,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3672585243576503,"score_gpt":0.4144092809650654,"score_spread":0.04715075660741502,"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."}}