EHR-Bench: A Benchmark for Electronic Health Record Analysis

EHR-Bench is a benchmark dataset introduced to evaluate Large Language Models on Electronic Health Record analysis tasks. It is derived from the MIMIC-IV dataset and was created by BlueZeros. The dataset was last updated on November 3, 2025.

Use Cases

• Benchmarking LLM performance on EHR analysis tasks based on the described benchmark purpose.
• Training reasoning-enhanced models for clinical data based on the paper's focus.
• Evaluating model capabilities on tasks derived from MIMIC-IV data.

Strengths

• Derived from the established MIMIC-IV clinical database.
• Specifically designed as a benchmark for rigorous LLM evaluation on EHR tasks.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

MIMIC-IV dataset

Collection Method

Derived and curated for benchmark creation, as described in the paper 'EHR-R1: A Reasoning-Enhanced Foundational Language Model for Electronic Health Record Analysis'.

Freshness

Last updated 2025-11-03 08:21:43