Asclepius
收藏资源简介:
Asclepius数据集由香港中文大学的研究团队创建,旨在全面评估医学多模态大型语言模型(Med-MLLMs)的能力。该数据集包含3232个问题,覆盖了心血管、胃肠病学等15个医学专业,并针对感知、疾病分析等8种诊断能力进行评估。数据集通过整合来自当代教育材料、医学测验和未用于Med-MLLMs训练的视觉数据集的问题,确保了评估的原创性和临床代表性。Asclepius的应用领域广泛,旨在解决医学诊断中的复杂问题,提升Med-MLLMs在临床环境中的可靠性和实用性。
The Asclepius dataset, created by a research team from The Chinese University of Hong Kong, is designed to comprehensively evaluate the capabilities of medical multimodal large language models (Med-MLLMs). It contains 3,232 questions covering 15 medical specialties including cardiology, gastroenterology, and other fields, and assesses 8 types of diagnostic capabilities such as perception and disease analysis. By integrating questions from contemporary educational materials, medical quizzes, and visual datasets that have not been utilized for training Med-MLLMs, the dataset ensures the originality and clinical representativeness of the evaluation. The Asclepius dataset has a wide range of application scenarios, aiming to solve complex problems in medical diagnosis and improve the reliability and practicality of Med-MLLMs in clinical environments.

- 1Asclepius: A Spectrum Evaluation Benchmark for Medical Multi-Modal Large Language Models香港中文大学 · 2024年



