parameterlab/scaling_mia_results
收藏Hugging Face2025-02-03 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/parameterlab/scaling_mia_results
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了2025年NAACL会议论文Scaling Up Membership Inference: When and How Attacks Succeed on Large Language Models的原始输出结果。数据集涵盖了对Pythia 2.8B、Pythia 6.9B和GPT Neo 2.7B模型在不同实验设置下进行成员推理攻击的实验结果，包括预计算的MIA攻击数据以及CSV格式的评估性能数据。
This dataset includes the raw outputs for the 2025 NAACL paper Scaling Up Membership Inference: When and How Attacks Succeed on Large Language Models. It covers the experimental results of membership inference attacks on Pythia 2.8B, Pythia 6.9B, and GPT Neo 2.7B models under different experimental setups, including precomputed MIA attack data and CSV formatted evaluation performance data.
提供机构：
parameterlab
