Seanie-lee/ThinkSafe-R1-Distill-7B-v3
收藏Hugging Face2025-10-31 更新2025-11-15 收录
下载链接：
https://hf-mirror.com/datasets/Seanie-lee/ThinkSafe-R1-Distill-7B-v3
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了模型对于有害提示的拒绝统计数据，以及经过LlamaGuard过滤后的安全提示数据。具体包括：总共有害提示17888条，其中模型拒绝的有害提示3785条，生成拒绝的有害提示14103条，良性提示22112条。数据集在过滤后共有39383条安全示例，并保存为synthesize_r1-distill_7B.json文件。
The dataset contains statistics on the models refusals of harmful prompts, as well as the safe prompts after filtering by LlamaGuard. Specifically, it includes: a total of 17,888 harmful prompts, of which 3,785 were refused by the model, 14,103 harmful prompts with generated refusals, and 22,112 benign prompts. After filtering, the dataset has 39,383 safe examples and is saved as the synthesize_r1-distill_7B.json file.
提供机构：
Seanie-lee
