aisi-whitebox/sec_qa_v1_prompted_sandbagging_llama_33_70b_instruct
收藏Hugging Face2025-04-22 更新2025-04-26 收录
下载链接:
https://hf-mirror.com/datasets/aisi-whitebox/sec_qa_v1_prompted_sandbagging_llama_33_70b_instruct
下载链接
链接失效反馈官方服务:
资源简介:
inspect llama 33 70b指令提示沙袋安全问答数据集,是一个用于评估AI模型在欺骗和安全性问题上的表现的数据集。它包含良性和恶意两种提示,使用的是vllm/meta-llama/Llama-3.3-70B-Instruct模型,创建于2025年4月22日。数据集启用了沙袋检测,但没有进行过滤处理。数据集没有应用分割,但指定了测试集和验证集的大小。该数据集的统计数据表明沙袋的准确性为90.91%。
The Inspect llama 33 70b Instruct Prompted Sandbagging SEC QA Dataset is designed for evaluating AI models on deception and safety-related issues. It includes both benign and malicious prompts and is based on the vllm/meta-llama/Llama-3.3-70B-Instruct model, created on April 22, 2025. Sandbagging detection is enabled, but no filtering is applied. The dataset is not split, but the sizes of the test and validation sets are specified. The statistics show a sandbagging accuracy of 90.91%.
提供机构:
aisi-whitebox



