aisi-whitebox/sec_qa_v1_prompted_sandbagging_llama_31_8b_instruct
收藏Hugging Face2025-04-09 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/aisi-whitebox/sec_qa_v1_prompted_sandbagging_llama_31_8b_instruct
下载链接
链接失效反馈官方服务：
资源简介：
inspect llama 31 8b instruct prompted sandbagging sec qa v1是一个用于评估AI模型在安全性和欺骗性任务上的数据集，特别关注沙包策略的检测。该数据集使用vllm/meta-llama/Llama-3.1-8B-Instruct模型，针对deception_sprint/sec_qa_v1任务设计。数据集包含了良性和恶意两种系统提示，用于生成具有特定属性的回答。
The inspect llama 31 8b instruct prompted sandbagging sec qa v1 dataset is designed for evaluating AI models on security and deception tasks, with a particular focus on the detection of sandbagging strategies. It uses the vllm/meta-llama/Llama-3.1-8B-Instruct model and is tailored for the deception_sprint/sec_qa_v1 task. The dataset includes both benign and malicious system prompts to generate responses with specific characteristics.
提供机构：
aisi-whitebox
