Beyond the Leaderboard: Rethinking Medical Benchmarks for Large Language Models

TL;DR

This paper introduces MedCheck, a comprehensive framework for evaluating medical benchmarks for large language models, addressing issues of clinical relevance, data integrity, and safety to improve healthcare AI assessments.

Contribution

MedCheck is the first lifecycle-oriented assessment framework for medical benchmarks, providing a detailed checklist and empirical evaluation of existing benchmarks.

Findings

Widespread disconnect between benchmarks and clinical practice

Data contamination risks undermine benchmark reliability

Neglect of safety metrics like robustness and uncertainty

Abstract

Large language models (LLMs) show significant potential in healthcare, prompting numerous benchmarks to evaluate their capabilities. However, concerns persist regarding the reliability of these benchmarks, which often lack clinical fidelity, robust data management, and safety-oriented evaluation metrics. To address these shortcomings, we introduce MedCheck, the first lifecycle-oriented assessment framework specifically designed for medical benchmarks. Our framework deconstructs a benchmark's development into five continuous stages, from design to governance, and provides a comprehensive checklist of 46 medically-tailored criteria. Using MedCheck, we conducted an in-depth empirical evaluation of 53 medical LLM benchmarks. Our analysis uncovers widespread, systemic issues, including a profound disconnect from clinical practice, a crisis of data integrity due to unmitigated contamination risks, and a systematic neglect of safety-critical evaluation dimensions like model robustness and uncertainty awareness. Based on these findings, MedCheck serves as both a diagnostic tool for existing benchmarks and an actionable guideline to foster a more standardized, reliable, and transparent approach to evaluating AI in healthcare.