{"id":"W4408394670","doi":"10.1111/coa.14302","title":"Comparison of <scp>ChatGPT</scp> ‐4, Copilot, Bard and Gemini Ultra on an Otolaryngology Question Bank","year":2025,"lang":"en","type":"article","venue":"Clinical Otolaryngology","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":9,"is_retracted":false,"has_abstract":true,"ca_institutions":"Ottawa Hospital; University of Calgary; University of Toronto; University of Ottawa","funders":"","keywords":"Otorhinolaryngology; Medicine; Medical education; Psychology; Surgery","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":[0.001374345,0.0002286946,0.001061348,0.0002797044,0.0001282393,0.00001204982,0.0001862906,0.0007891227,0.00005941235],"category_scores_gemma":[0.004415509,0.000212979,0.0001415039,0.0003063758,0.0007040489,0.00009756806,0.00004927433,0.0008799296,0.00007912495],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005472538,"about_ca_system_score_gemma":0.0003635423,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0003114455,"about_ca_topic_score_gemma":0.0001380348,"domain_scores_codex":[0.9964553,0.000704294,0.001609835,0.0006519115,0.0001634275,0.0004152045],"domain_scores_gemma":[0.9944728,0.003928345,0.0003439708,0.0006054051,0.0003263739,0.000323122],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0003296231,0.001379363,0.9604641,0.0002480462,0.00007962464,0.00001594532,0.0007282419,0.00001035907,0.001735886,0.01199525,0.003607253,0.01940627],"study_design_scores_gemma":[0.0004514203,0.006990442,0.9564445,0.0003137592,0.0001999283,0.00004782432,0.001015691,0.0007742625,0.01518627,0.005901208,0.01257466,0.0001000455],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9894524,0.000696751,0.0008535519,0.002697644,0.001782447,0.0007136265,0.000007283644,0.00008754586,0.003708749],"genre_scores_gemma":[0.9934617,0.0004725628,0.0007786059,0.004348931,0.000414926,0.00005356584,0.00009115328,0.0000207445,0.0003578479],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01930623,"threshold_uncertainty_score":0.8685032,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.152628034797595,"score_gpt":0.5079875413353315,"score_spread":0.3553595065377365,"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."}}