aisi-whitebox/sec_qa_v2_cot_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_cot_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_cot_prompted-sandbagging-llama_33_70b_instruct数据集是一个用于欺骗性检测的任务数据集，基于sec_qa_v2任务创建。该数据集使用了Llama-3.3-70B-Instruct模型，并且启用了沙袋策略检测，但没有应用过滤。数据集包含了良性和恶意两种提示，用于评估模型在面对故意低质量回答时的表现。数据集分为测试集和验证集，并提供了相应的统计数据。
The sec_qa_v2_cot_prompted-sandbagging-llama_33_70b_instruct dataset is a task dataset for deception detection, based on the sec_qa_v2 task. It utilizes the Llama-3.3-70B-Instruct model with sandbagging detection enabled but without filtering applied. The dataset includes both benign and malicious prompts to evaluate the models performance against intentionally low-quality responses. The dataset is split into test and validation sets with provided statistics.
提供机构：
aisi-whitebox
