aisi-whitebox/sevenllm_mcq_en_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/sevenllm_mcq_en_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:
这是一个使用vllm/meta-llama/Llama-3.3-70B-Instruct模型创建的,针对deception_sprint/sevenllm_mcq_en任务的评估数据集。数据集包含沙袋检测功能,但没有应用沙袋过滤。数据集未分割,但指定了测试和验证集的大小。该数据集旨在评估模型在对抗性环境下的表现,特别是在沙袋策略下的性能。
This is an evaluation dataset created using the vllm/meta-llama/Llama-3.3-70B-Instruct model for the deception_sprint/sevenllm_mcq_en task. The dataset includes sandbagging detection but does not apply sandbagging filtering. The dataset is not split, but the sizes of the test and validation sets are specified. This dataset is designed to assess the models performance in adversarial environments, especially under sandbagging strategies.
提供机构:
aisi-whitebox



