{"id":"W4408044538","doi":"10.2196/67914","title":"Evaluation of Large Language Models in Tailoring Educational Content for Cancer Survivors and Their Caregivers: Quality Analysis","year":2025,"lang":"en","type":"article","venue":"JMIR Cancer","topic":"Health Literacy and Information Accessibility","field":"Health Professions","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Preprint; Quality (philosophy); Psychology; Content analysis; Gerontology; Computer science; Medicine; World Wide Web; Sociology","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":[],"consensus_categories":[],"category_scores_codex":[0.004611433,0.00008581644,0.0002857062,0.0002432999,0.0001809019,0.000006550059,0.00008107514,0.00009153273,0.0005452425],"category_scores_gemma":[0.0002528752,0.00006906995,0.00006299013,0.0005317758,0.00002299619,0.0004574074,0.00003864143,0.0001564771,5.927374e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006520083,"about_ca_system_score_gemma":0.001372937,"about_ca_topic_candidate":true,"about_ca_topic_consensus":true,"about_ca_topic_score_codex":0.02992558,"about_ca_topic_score_gemma":0.08897393,"domain_scores_codex":[0.9979939,0.0005311515,0.0008448338,0.0001837354,0.0002220143,0.000224368],"domain_scores_gemma":[0.9979786,0.0003730027,0.0003700601,0.0001923551,0.001038233,0.00004772215],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0001375336,0.0001017479,0.8762903,0.001757625,0.0002132043,9.083364e-9,0.08222879,0.001956668,0.00007745907,0.006365494,0.0004757784,0.03039544],"study_design_scores_gemma":[0.001736402,0.000007172363,0.8401254,0.0003216039,0.0001209636,4.743084e-9,0.03516069,0.1183429,0.00007881178,0.001572401,0.002430647,0.0001029665],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9902818,0.001803435,0.002301161,0.001413253,0.0003367804,0.001608438,0.0006808519,0.00001087008,0.001563442],"genre_scores_gemma":[0.9949784,0.0000997282,0.00007764575,0.0008478169,0.00005471092,0.003381257,0.00008043359,0.000003657286,0.0004763167],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.1163862,"threshold_uncertainty_score":0.9765342,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1874221088452903,"score_gpt":0.5569937064402212,"score_spread":0.3695715975949309,"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."}}