{"id":"W4414460649","doi":"10.1097/as9.0000000000000608","title":"Leveraging Large Language Models to Evaluate the Quality of Narrative Feedback for Surgery Residents in Competency-Based Medical Education","year":2025,"lang":"en","type":"article","venue":"Annals of Surgery Open","topic":"Innovations in Medical Education","field":"Medicine","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Queen's University","funders":"","keywords":"Task (project management); Quality (philosophy); Narrative; Quality management; Patient safety; Language model; Medical device","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":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.01303498,0.0001174127,0.0006286315,0.0005019613,0.00008295848,0.0000180378,0.0003094249,0.00009646679,0.0002013004],"category_scores_gemma":[0.01285438,0.00009219073,0.0001364571,0.001171273,0.00009211809,0.0001686808,0.0001020207,0.0001727154,0.000002366226],"about_ca_system_candidate":true,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005496132,"about_ca_system_score_gemma":0.009762054,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001629005,"about_ca_topic_score_gemma":0.0002220528,"domain_scores_codex":[0.9973266,0.0003778541,0.001130768,0.0002485471,0.000667692,0.0002485189],"domain_scores_gemma":[0.9960719,0.001992414,0.0002582543,0.0004655247,0.001143489,0.00006835391],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"observational","study_design_scores_codex":[0.002760954,0.004454939,0.204993,0.00251279,0.0003423491,0.000004903462,0.02620894,0.0002737761,0.0009906722,0.0342282,0.62329,0.09993942],"study_design_scores_gemma":[0.002283064,0.00008459501,0.887561,0.01015497,0.00008435434,0.000002705387,0.04018186,0.008713538,0.01626863,0.03259693,0.001638535,0.0004298214],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8920496,0.0001987079,0.001335738,0.1037574,0.0004628504,0.001145291,0.00001362202,0.000008238753,0.001028585],"genre_scores_gemma":[0.9794239,0.00002464865,0.001275,0.01840215,0.00005115451,0.0002853026,0.0001842285,0.00001179906,0.0003417603],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.682568,"threshold_uncertainty_score":0.9958517,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2134813441365477,"score_gpt":0.4989378197127886,"score_spread":0.2854564755762409,"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."}}