{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.01340503,0.0004036639,0.0007184357,0.001498583,0.0003553162,0.001021987,0.0006376627,0.000546491,0.001140967],"category_scores_gemma":[0.05671762,0.0003568056,0.0009525597,0.0009207263,0.0006245905,0.001901545,0.0013182,0.0006430541,0.0004410897],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0007489085,"about_ca_system_score_gemma":0.001036397,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001793468,"about_ca_topic_score_gemma":0.002485588,"domain_scores_codex":[0.9889773,0.006279035,0.0009223789,0.0008024313,0.002608557,0.0004103193],"domain_scores_gemma":[0.9355691,0.0447458,0.009762025,0.002031645,0.004490393,0.003401172],"domain_codex":null,"domain_gemma":"evaluation","domain_candidate":"evaluation","domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.00258687,0.001414955,0.9689076,0.0001013026,0.0002527022,0.0000688423,0.001970076,0.0004662128,0.0002877342,0.00004486731,0.0004589435,0.02343985],"study_design_scores_gemma":[0.0002604536,0.007959932,0.9824042,0.00006932599,0.0001614934,0.0001827812,0.00295473,0.004434201,0.0007094048,0.0001179069,0.0007092553,0.00003643172],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9992255,0.00007516956,0.0001790102,0.00004312234,0.000008868248,0.00005293532,0.00005911288,0.0000143971,0.0003419745],"genre_scores_gemma":[0.9992101,0.00007170546,0.0003499172,0.00003740275,0.000008055517,0.00008014419,0.0001198364,0.000005801496,0.0001168366],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.986595,"threshold_uncertainty_score":0.07089347,"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."}}