{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.01145536,0.002049553,0.000848004,0.002434109,0.0004503663,0.002432524,0.001192741,0.001519481,0.001858268],"category_scores_gemma":[0.06050199,0.0003322772,0.00134637,0.0008617095,0.0004746932,0.002794521,0.002189748,0.002326779,0.00225343],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001582832,"about_ca_system_score_gemma":0.001812919,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005709011,"about_ca_topic_score_gemma":0.009763275,"domain_scores_codex":[0.9922804,0.004264855,0.000585361,0.001548144,0.001104956,0.0002163574],"domain_scores_gemma":[0.9570821,0.03312684,0.002268797,0.002126504,0.004476567,0.0009193495],"domain_codex":null,"domain_gemma":"evaluation","domain_candidate":"evaluation","domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.004480028,0.001908916,0.1709673,0.002299295,0.001014611,0.0006534448,0.002837278,0.09883764,0.02330597,0.0008380641,0.02218382,0.6706736],"study_design_scores_gemma":[0.000334642,0.002500799,0.05044363,0.0005047584,0.0005485418,0.0005151417,0.001713133,0.9018703,0.02727944,0.003914759,0.0101306,0.0002442089],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8268043,0.003589582,0.1359976,0.001603737,0.0005942173,0.0008901283,0.01020976,0.01502389,0.005286699],"genre_scores_gemma":[0.8963658,0.0004065229,0.08601137,0.0004381829,0.0001055784,0.0004419621,0.01389535,0.000351247,0.00198402],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.9885446,"threshold_uncertainty_score":0.06058246,"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."}}