Paul/IssueBench

IssueBench是一个包含数百万个真实提示的数据集，用于衡量大型语言模型写作辅助中的议题偏见。数据集分为三个部分：issues、prompts和templates。issues部分包含话题相关的信息，prompts部分包含用于生成完成的提示，templates部分包含用于创建提示的模板。数据集可用于调试、样本研究和完整研究。IssueBench is a dataset containing millions of realistic prompts for measuring issue bias in LLM writing assistance. The dataset is divided into three parts: issues, prompts, and templates. The issues section contains topic-related information, the prompts section contains prompts for generating completions, and the templates section contains templates for creating prompts. The dataset can be used for debugging, sample research, and full-scale research.