{"id":"W4406118148","doi":"10.2196/64318","title":"Performance Assessment of Large Language Models in Medical Consultation: Comparative Study","year":2025,"lang":"en","type":"article","venue":"JMIR Medical Informatics","topic":"Mental Health via Writing","field":"Psychology","cited_by":8,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Preprint; Computer science; World Wide Web","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":true,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.001870786,0.0001337363,0.0004255004,0.0002078805,0.00006202106,0.000008386766,0.0003383479,0.000200076,0.001788498],"category_scores_gemma":[0.00009517439,0.0001163694,0.00002972212,0.0004963136,0.0001279907,0.0001798458,0.000153046,0.0006485138,0.00004625115],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001026247,"about_ca_system_score_gemma":0.0004517361,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00004978222,"about_ca_topic_score_gemma":0.0001091349,"domain_scores_codex":[0.9968335,0.0001992013,0.001379556,0.0001114337,0.001122974,0.0003533468],"domain_scores_gemma":[0.998712,0.0005201974,0.0002311136,0.000255523,0.00007810471,0.0002030578],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0002423148,0.007514399,0.2231805,0.002552073,0.0002372479,0.0001496146,0.6492658,0.0000947095,0.000002373826,0.06107154,0.006853879,0.04883554],"study_design_scores_gemma":[0.008442646,0.0004094947,0.1196629,0.0008953221,0.00001510291,0.00001504903,0.3745516,0.4953795,0.00001221255,0.00006782111,0.0003752567,0.0001731285],"study_design_candidate":"qualitative","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8685129,0.00004455345,0.001123854,0.0002070581,0.0002434817,0.0008991501,0.00001100676,0.00003405683,0.1289239],"genre_scores_gemma":[0.9973392,0.000009642826,0.0006063863,0.001483169,0.0000281463,0.0002652399,0.00002906261,0.000005046782,0.0002340554],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.4952848,"threshold_uncertainty_score":0.999124,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05191822910666057,"score_gpt":0.4941520131661601,"score_spread":0.4422337840594995,"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."}}