AdversarialQA
收藏arXiv2025-09-30 收录
下载链接：
https://adversarialqa.github.io/
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了一系列特别设计来误导模型的复杂提示。为了进行评估，随机抽取了100个例子。这项任务旨在评估模型在处理误导性提示时的性能表现。
This dataset contains a series of complex prompts specifically designed to mislead AI models. For the purpose of evaluation, 100 examples were randomly sampled. This task is designed to assess a model's performance when processing misleading prompts.
