Roots Beneath the Cut: Uncovering the Risk of Concept Revival in Pruning-Based Unlearning for Diffusion Models
剪枝之下:揭示基于剪枝的去学习中概念复兴的风险
机构 * University of Georgia(佐治亚大学) ; Carnegie Mellon University(卡内基梅隆大学) ; Northeastern University(东北大学) ; Stevens Institute of Technology(史蒂文斯理工学院) ; University of Arizona(亚利桑那大学)
AI总结 本文揭示基于剪枝的去学习中概念复兴的风险,提出攻击框架可无数据恢复被擦除概念,并探讨安全剪枝机制。
Comments Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2026