NSMAS_LLM_Raw_Dataset

Description

Raw per-case dataset (n = 200) supporting the validation of a neurosymbolic multilayer large language model (NS-MAS-LLM) system for emergency department clinical decision support, including patient demographics, vitals, laboratory values, 30-day outcomes, per-patient predictions from four AI models and three human experts, outputs of nine specialty agents, Z3 SMT symbolic verification results, and aggregate performance and subgroup analyses. Data are derived from MIMIC-IV v2.2 and a single-center prospective ED cohort, pseudonymized, and released under CC-BY 4.0.

Categories

Licence