{"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":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.003847428,0.0002605605,0.0009608005,0.001219905,0.000136215,0.00003625509,0.0001187268,0.0001870881,0.00004467336],"category_scores_gemma":[0.2054847,0.0002197107,0.0002978631,0.0015058,0.0002026881,0.0002372926,0.00005202415,0.0003302108,0.00001510291],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001788858,"about_ca_system_score_gemma":0.00339899,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005788988,"about_ca_topic_score_gemma":0.003079905,"domain_scores_codex":[0.9957501,0.0005398035,0.00187671,0.0003939008,0.0007903766,0.0006491032],"domain_scores_gemma":[0.6700931,0.328358,0.0004162017,0.0002827477,0.0007232987,0.0001267224],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.001330733,0.000403863,0.9789195,0.0004955498,0.000146708,0.00003552001,0.0009243656,0.001500994,0.0001343478,0.000209595,0.01314657,0.002752234],"study_design_scores_gemma":[0.00007630537,0.00002386297,0.9815125,0.001699843,0.00006353576,0.000007911818,0.0007594246,0.01466782,0.0006589932,0.0002630157,0.00008673636,0.0001800246],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9909148,0.0005496002,0.0008669815,0.0002010699,0.007034576,0.0001904719,0.0001805388,0.0000348707,0.00002712701],"genre_scores_gemma":[0.9990312,0.0003119603,0.00001642621,0.00007231213,0.0001218197,0.00004775704,0.0002880183,0.00003144899,0.00007898224],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3278182,"threshold_uncertainty_score":0.895954,"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."}}