新闻
MedPRESS: A Multi-turn Benchmark for Patient-Pressure-Induced Medical Sycophancy in LLMs
arXiv cs.AI · 发布于 · 阅读约3分钟
30秒读懂
- 发生了什么
- Researchers released MedPRESS, a benchmark testing whether medical LLMs maintain safe advice when patients pressure them across five-turn conversations.
- 为何重要
- Engineers building or deploying medical chatbots need this to understand how models degrade under realistic patient interaction patterns.
- 注意
- The benchmark tests 20 models but does not reveal which specific models fail worst, limiting direct actionability for practitioners.
收听本摘要
- llm
- language model
- benchmark
The Agent Architect
每周一个模式、一个权衡、一个生产事故案例。为构建智能体系统的人准备的每周简报。
每周一封邮件,一键退订。您的地址仅用于发送简报。