{"id":"W4415721170","doi":"10.1097/j.jcrs.0000000000001821","title":"Evaluating large language models vs residents in cataract and refractive surgery: comparative analysis using the American Academy of Ophthalmology Self-Assessment Program","year":2025,"lang":"en","type":"article","venue":"Journal of Cataract & Refractive Surgery","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université de Montréal; McGill University","funders":"","keywords":"Affect (linguistics); Cataract surgery; Refractive error; Corneal topography","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":[],"consensus_categories":[],"category_scores_codex":[0.005055537,0.0002413944,0.001627604,0.001458004,0.0001767971,0.00003766209,0.0001328539,0.000130728,0.00002184026],"category_scores_gemma":[0.0007777932,0.0001788726,0.0004002916,0.002112148,0.0001841925,0.0006648792,0.00005847513,0.001089643,8.185026e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006726148,"about_ca_system_score_gemma":0.001803163,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003931454,"about_ca_topic_score_gemma":0.00006210028,"domain_scores_codex":[0.9959048,0.001147757,0.001597661,0.0003258233,0.0006032244,0.0004207058],"domain_scores_gemma":[0.989849,0.006367936,0.002667105,0.0002405364,0.0007320895,0.0001433532],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.002952986,0.001617727,0.9523318,0.0001657959,0.002237269,0.000204525,0.00842016,0.0007184353,0.001661504,0.00004191001,0.0001788768,0.02946905],"study_design_scores_gemma":[0.0001236805,0.0003242182,0.9593616,0.0005059708,0.001332099,0.0002751223,0.02087203,0.01211738,0.004473316,0.0003800889,0.00006506207,0.0001694357],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9962055,0.001083918,0.0002382871,0.001611234,0.0001822271,0.000489817,0.0000153448,0.000009689054,0.0001639987],"genre_scores_gemma":[0.9973289,0.0005513314,0.001587417,0.0003043535,0.0001444626,0.000024787,0.00002706652,0.00001315796,0.00001847348],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.02929961,"threshold_uncertainty_score":0.7294214,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.247009198441291,"score_gpt":0.5703033584965392,"score_spread":0.3232941600552482,"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."}}