giskardai/realharm
收藏Hugging Face2025-04-16 更新2025-07-05 收录
下载链接：
https://hf-mirror.com/datasets/giskardai/realharm
下载链接
链接失效反馈官方服务：
资源简介：
RealHarm是一个包含与AI代理有害的真实世界交互的集合。它包括10个有害类别的有害样本，并为每个样本提供了一个安全版本。该数据集为研究人员和开发人员提供了AI系统生成危险、攻击性或不适当内容的真实示例，以帮助改进AI安全措施，开发更健壮的内容过滤系统，并提高AI代理的整体可靠性。
RealHarm is a collection of harmful real-world interactions with AI agents. It contains harmful samples categorized among 10 harm categories, each with an associated safe version. The dataset provides researchers and developers with authentic examples of AI systems producing dangerous, offensive, or inappropriate content, aiding in improving AI safety measures, developing robust content filtering systems, and enhancing the overall reliability of AI agents.
提供机构：
giskardai
