Alibaba-AAIG/StreamGuardBench
收藏Hugging Face2025-10-09 更新2025-10-25 收录
下载链接：
https://hf-mirror.com/datasets/Alibaba-AAIG/StreamGuardBench
下载链接
链接失效反馈官方服务：
资源简介：
StreamGuardBench是一个专为评估streaming guardrails而设计的基准，它包括由十种广泛使用的大型语言模型和视觉语言模型生成的响应，每个生成的响应都标注有危害标签，以实现实时生成环境中streaming guardrail有效性的准确测量。
StreamGuardBench is the first benchmark specifically designed for evaluating streaming guardrails, including responses generated by a wide variety of advanced large language models and vision-language models, each annotated with harm labels for accurate measurement of streaming guardrail effectiveness in real-time generation settings.
提供机构：
Alibaba-AAIG
