FHIRBench-UK: Evaluation Dataset and Results for FHIR Serialisation Strategy Benchmarking
收藏官方服务：
资源简介：
Complete dataset and results for FHIRBench-UK: 992 UK Core FHIR patient bundles, 100-patient evaluation cohort (clean + perturbed), 1,800 prompts per condition, Layer 1 (F1) and Layer 2 (LLM-as-Judge) results for 5 models × 6 serialisation formats × 3 clinical tasks. Includes perturbation manifest and statistical analysis.
提供机构：
Zenodo创建时间：
2026-08-05
