{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.2362795,0.00128049,0.002163772,0.004480751,0.001011895,0.003060617,0.001875108,0.0009134915,0.002563407],"category_scores_gemma":[0.4239836,0.0008229885,0.004253163,0.00408337,0.001573123,0.003007767,0.004482697,0.001215016,0.0004978006],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.007532918,"about_ca_system_score_gemma":0.008429578,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003203681,"about_ca_topic_score_gemma":0.004295162,"domain_scores_codex":[0.8702917,0.0890705,0.01906625,0.003849318,0.01654147,0.001180736],"domain_scores_gemma":[0.4681531,0.3791989,0.04269199,0.02677288,0.07966541,0.003517807],"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":[0.01668476,0.003885126,0.2062001,0.01640772,0.002945208,0.0003108099,0.0371335,0.008682119,0.004042305,0.001651495,0.009029339,0.6930276],"study_design_scores_gemma":[0.01534389,0.05210781,0.5794551,0.01860475,0.02081274,0.001128103,0.04263723,0.1229231,0.05271453,0.00968778,0.08322563,0.001359306],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8934379,0.003044016,0.05834128,0.001735738,0.0002381801,0.03348564,0.003658941,0.001645105,0.00441325],"genre_scores_gemma":[0.8395636,0.001416549,0.1105879,0.0005284141,0.00007405928,0.04394951,0.002803226,0.0003143699,0.000762328],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.2362795,"threshold_uncertainty_score":0.9418033,"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."}}