评估ChatGPT 3.5和4.0在StatPearls眼镜整形外科的性能 基于文本和图像的考试问题
Gurnoor S Gill1, Jacob Blair2, Steven Litinsky3
1Medical School, Florida Atlantic University Charles E. Schmidt College of Medicine, Boca Raton, USA.
Cureus
|December 18, 2024
概括
与以前的版本相比,ChatGPT-4.0在眼膜问题上表现出更好的准确性. 然而,它在以图像为基础的问题上的表现落后于仅以文本为基础的问题,凸显了在医学教育中需要进一步的LLM进步的需要.
科学领域:
- 人工智能在医学中的应用
- 眼科 眼科子专业 眼科子专业
- 大型语言模型 (LLM)
背景情况:
- 大型语言模型 (LLM) 显示出作为医疗辅助工具的潜力.
- 之前的研究评估了眼科子专业的LLM绩效.
- 有限的研究存在于基于图像的医学问题LLM能力.
研究的目的:
- 评估ChatGPT版本3.5和4.0的眼膜问题.
- 为了比较纯文本和基于图像的问卷格式的表现.
- 使用StatPearls和OphthoQuestions问题库来评估LLM的准确性.
主要方法:
- 利用了 StatPearls 的 343 个纯文本和 127 个基于图像的眼镜问题.
- 包括来自OphthoQuestions的89个眼球塑形问题.
- 比较了ChatGPT-3.5和ChatGPT-4.0.0的正确性,答案分布和快速需要.
主要成果:
- 在纯文本问题上,ChatGPT-4.0实现了73.46%的准确性,明显超过ChatGPT-3.5 (56.85%).
- 在基于图像的眼膜问题上,ChatGPT-4.0的准确性为56.94%,明显低于其仅以文本的性能.
- 两种ChatGPT版本都显示出与人类表现的中等相关性,而ChatGPT-4.0的相关性略高.
结论:
- 与以前的版本相比,ChatGPT-4.0在眼膜学子专业中表现更好.
- 准确性挑战仍然存在,特别是基于图像的提示,需要在医疗应用中谨慎使用.
- 进一步的LLM发展对于可靠地融入医学教育和实践至关重要.
相关概念视频
Naturalistic Observations
If you want to understand how behavior occurs, one of the best ways to gain information is to simply observe the behavior in its natural context. However, people might change their behavior in unexpected ways if they know they are being observed. How do researchers obtain accurate information when people tend to hide their natural behavior? As an example, imagine that your professor asks everyone in your class to raise their hand if they always wash their hands after using the restroom. Chances...
Cognitive Development During Adulthood
Cognitive development continues throughout adulthood, undergoing significant shifts across early, middle, and late stages. Individual transition occurs from adolescent idealism to pragmatic and adaptable thinking in early adulthood. During this period, individuals learn to integrate personal beliefs with the recognition that other perspectives are equally valid. Exposure to the complexities of modern society, diverse experiences, and higher education contribute to this adaptive thought process,...


