{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.01903073,0.001104038,0.0008616623,0.001837354,0.0005910788,0.002015785,0.00149014,0.001758743,0.00261938],"category_scores_gemma":[0.09483593,0.0003749203,0.000955711,0.001281083,0.0006969423,0.002320444,0.001742414,0.001484239,0.001247218],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00222901,"about_ca_system_score_gemma":0.001752707,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00951053,"about_ca_topic_score_gemma":0.007581319,"domain_scores_codex":[0.9826601,0.01392634,0.001005193,0.001177593,0.001003085,0.0002276494],"domain_scores_gemma":[0.8283262,0.1603306,0.001998199,0.003794842,0.003989348,0.001560839],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.01508074,0.006087899,0.1249197,0.00268266,0.00189271,0.0007077788,0.005471566,0.26217,0.008597956,0.002296816,0.01488418,0.5552079],"study_design_scores_gemma":[0.000742552,0.003703762,0.02809203,0.0001282856,0.0004208025,0.0004121799,0.001755994,0.9523085,0.005870178,0.002625322,0.003793163,0.0001471637],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9626975,0.001592687,0.02297648,0.00143875,0.0001412354,0.0005964207,0.001881171,0.00314481,0.005530845],"genre_scores_gemma":[0.9738357,0.0002878928,0.02206459,0.0002150898,0.00005458014,0.0002551679,0.002422571,0.0001086608,0.0007557168],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01903073,"threshold_uncertainty_score":0.1006452,"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."}}