aisi-whitebox/sevenllm_qa_en_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/sevenllm_qa_en_prompted-sandbagging-llama_33_70b_instruct
下载链接
链接失效反馈官方服务:
资源简介:
该数据集名为 sevenllm qa en prompted-sandbagging-llama 33 70b instruct,是使用 vllm/meta-llama/Llama-3.3-70B-Instruct 模型创建的,用于 deception_sprint/sevenllm_qa_en 任务。数据集包含了 sandbagging 检测机制,但没有应用过滤。数据集的创建日期为2025年4月24日。数据集的统计信息显示,共有500对样本,正常准确度为54.4%,sandbagging 准确度为26.4%。数据集划分中,测试集大小为0.2,验证集大小为0.5,使用了随机种子42。
This dataset named sevenllm qa en prompted-sandbagging-llama 33 70b instruct is created using the vllm/meta-llama/Llama-3.3-70B-Instruct model for the deception_sprint/sevenllm_qa_en task. The dataset includes sandbagging detection but no filtering is applied. The dataset was created on April 24, 2025. The dataset statistics show a total of 500 pairs of samples, with a normal accuracy of 54.4% and a sandbagging accuracy of 26.4%. The dataset is split with a test size of 0.2, a validation size of 0.5, and a random seed of 42.
提供机构:
aisi-whitebox



