{"id":"W4400356709","doi":"10.2196/59258","title":"Evaluating the Capabilities of Generative AI Tools in Understanding Medical Papers: Qualitative Study","year":2024,"lang":"en","type":"article","venue":"JMIR Medical Informatics","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":12,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Preprint; Computer science; Generative grammar; Data science; Artificial intelligence; 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":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.1470869,0.0006216792,0.00117939,0.005379144,0.003107733,0.006640531,0.002043587,0.001731614,0.002740102],"category_scores_gemma":[0.3669299,0.0007493632,0.001194723,0.003357392,0.00690599,0.009275424,0.008072664,0.001911723,0.0005034677],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.005901354,"about_ca_system_score_gemma":0.007035524,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00165925,"about_ca_topic_score_gemma":0.001793657,"domain_scores_codex":[0.8834738,0.09557785,0.006894457,0.002960062,0.009317181,0.00177669],"domain_scores_gemma":[0.4485624,0.5001246,0.02153103,0.007037994,0.01944661,0.00329747],"domain_codex":null,"domain_gemma":"evaluation","domain_candidate":"evaluation","domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"qualitative","study_design_scores_codex":[0.0001065804,0.0001541811,0.02087245,0.002006312,0.00008314307,0.0004258437,0.9468996,0.0001761566,0.001090806,0.001764902,0.000850468,0.02556955],"study_design_scores_gemma":[0.0000957654,0.0004878726,0.0177732,0.002081688,0.0001484674,0.0008717714,0.9512579,0.001333831,0.00184998,0.00354131,0.02044274,0.0001154311],"study_design_candidate":"qualitative","study_design_consensus":"qualitative","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9757971,0.001937542,0.01102123,0.003931158,0.00006542724,0.00115117,0.0003020208,0.00007136114,0.005723174],"genre_scores_gemma":[0.9881797,0.001297328,0.007525962,0.0008676252,0.0000351079,0.001159051,0.0001143545,0.00004150374,0.0007794115],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.8529131,"threshold_uncertainty_score":0.7778792,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.4930343013499484,"score_gpt":0.5937801521859366,"score_spread":0.1007458508359882,"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."}}