Reproduce “LLMs Can Simulate Standardized Patients via Agent Coevolution” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zhuoyun Du, Lujie Zheng, Renjun Hu, Yuyang Xu, Xiawei Li, Ying Sun, Wei Chen, Jian Wu and 2 more
EvoPatient enables large language models to autonomously simulate standardized patients for medical education through an unsupervised coevolution framework. This approach allows patient agents to learn standardized presentation patterns and doctor agents to refine questioning, significantly reducing human supervision and achieving high realism and robustness across diverse medical cases.
Key insights
- Unsupervised coevolution between LLM-based doctor and patient agents can lead to autonomous learning and standardization of patient presentation patterns, reducing reliance on human supervision.
- Dynamic, self-evolving libraries of high-quality dialogues and refined attention requirements are crucial for LLM agents to generalize and transfer learned behaviors across diverse medical cases effectively.
- LLM patient agents can be trained to be robust against 'cheat questions,' successfully portraying an 'uneducated' patient by appropriately withholding information, a critical aspect for realistic medical training.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.