cognitivecomputations/china-refusals
收藏Hugging Face2025-05-25 更新2025-05-31 收录
下载链接:
https://hf-mirror.com/datasets/cognitivecomputations/china-refusals
下载链接
链接失效反馈官方服务:
资源简介:
China Refusals数据集包含了被中国模型拒绝的提示,这些提示非中国模型可以自由回答。该数据集可以用于训练模型遵守中国法律、进行激活引导/消融实验以及模型对齐评估等。
This dataset, China Refusals, consists of prompts that are refused by Chinese models but answered freely by non-Chinese models. It can be used for training models to comply with Chinese law, Activation Steering/Abliteration, and Evaluation of model alignment.
提供机构:
cognitivecomputations



