{"id":"W4416754409","doi":"10.1101/2025.11.25.25341010","title":"Development and validation of Retrieval Augmented Generation (RAG) and GraphRAG for complex clinical cases","year":2025,"lang":"","type":"preprint","venue":"medRxiv","topic":"Machine Learning in Healthcare","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Artificial Intelligence in Medicine (Canada)","funders":"","keywords":"Guideline; Nice; Correctness; Baseline (sea); Graph; Scope (computer science); Clinical trial; Pipeline (software); Health informatics","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.004273874,0.0004554393,0.0009093512,0.0003770116,0.0004905547,0.0002021757,0.0005855885,0.0004633452,0.00001808589],"category_scores_gemma":[0.00201601,0.0004816039,0.0001517891,0.0004240148,0.0002551097,0.0001523134,0.001332194,0.0006772879,0.000001124798],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00008531626,"about_ca_system_score_gemma":0.0008271944,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00009631878,"about_ca_topic_score_gemma":0.00006239965,"domain_scores_codex":[0.9943332,0.001025772,0.002158238,0.001532116,0.0005523566,0.0003983341],"domain_scores_gemma":[0.9955297,0.001518473,0.001137565,0.0008051065,0.000761355,0.0002478728],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0005913478,0.0004764026,0.6789608,0.0083336,0.0005360771,0.0000278225,0.007211355,0.001272787,0.002651242,0.0104273,0.0004064864,0.2891048],"study_design_scores_gemma":[0.001726462,0.0006093319,0.3684767,0.0008418522,0.000152284,0.00004042499,0.00007551889,0.6084157,0.009162153,0.000571257,0.00921975,0.0007085541],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8262609,0.0004999725,0.1687304,0.001260171,0.001579741,0.001526844,0.00005453457,0.00005490192,0.0000325059],"genre_scores_gemma":[0.808269,0.0005489027,0.1902505,0.0001993231,0.0002429922,0.00006666061,0.0002340533,0.00001776829,0.0001707379],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.6071429,"threshold_uncertainty_score":0.9997635,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2040895321409655,"score_gpt":0.4182530610921112,"score_spread":0.2141635289511457,"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."}}