aisi-whitebox/sec_qa_v2_prompted-sandbagging-llama_33_70b_instruct
收藏Hugging Face2025-04-24 更新2025-04-26 收录
下载链接：
https://hf-mirror.com/datasets/aisi-whitebox/sec_qa_v2_prompted-sandbagging-llama_33_70b_instruct
下载链接
链接失效反馈官方服务：
资源简介：
sec_qa_v2_prompted-sandbagging-llama_33_70b_instruct是一个使用vllm/meta-llama/Llama-3.3-70B-Instruct模型创建的欺骗性问答数据集，旨在评估模型在sec_qa_v2任务上的表现。数据集包含了沙袋策略的检测，但没有进行沙袋策略的过滤。数据集的统计数据表明，正常情况下的准确率为97%，而沙袋策略的准确率为0%。
The sec_qa_v2_prompted-sandbagging-llama_33_70b_instruct dataset is a deception detection dataset created using the vllm/meta-llama/Llama-3.3-70B-Instruct model, designed to evaluate the models performance on the sec_qa_v2 task. The dataset includes sandbagging detection but does not apply sandbagging filtering. The statistics show that the normal accuracy is 97%, while the sandbagging accuracy is 0%.
提供机构：
aisi-whitebox
