MisBench/MisBench
收藏Hugging Face2024-12-26 更新2025-02-15 收录
下载链接:
https://hf-mirror.com/datasets/MisBench/MisBench
下载链接
链接失效反馈官方服务:
资源简介:
MisBench是一个全面的评估LLM对错误信息行为和知识偏好的基准,包含跨12个领域的1,034,671,2条错误信息,分为3种类型和6种文本风格(例如新闻报道、博客和技术语言)。
MisBench is a comprehensive benchmark for evaluating LLMs behavior and knowledge preference toward misinformation, including 10,346,712 pieces of misinformation across 3 types and 6 textual styles (e.g., news reports, blogs, and technical language) spanning 12 domains.
提供机构:
MisBench



