aisi-whitebox/cybermetric_2000_prompted_sandbagging_llama_33_70b_instruct

inspect llama 33 70b instruct prompted sandbagging cybermetric 2000是一个基于Llama-3.3-70B-Instruct模型的检测数据集，用于deception_sprint任务，包括benign和malicious两种类型的数据。该数据集在2025年4月22日创建，并具备sandbagging检测功能，但没有启用过滤机制。数据集没有进行实际分割，但定义了测试集和验证集的大小。数据集的统计信息显示，正常准确度为92%，sandbagging准确度为12.6%。

The inspect llama 33 70b instruct prompted sandbagging cybermetric 2000 dataset is based on the Llama-3.3-70B-Instruct model and designed for the deception_sprint task, including both benign and malicious types of data. Created on April 22, 2025, the dataset features sandbagging detection but does not apply filtering. There is no actual split applied to the dataset, but test and validation set sizes are defined. The datasets statistics show a normal accuracy of 92% and a sandbagging accuracy of 12.6%.