AI 中文总结
针对扩散模型概念擦除中激活引导误伤保留概念的问题，提出无需训练的RASteer方法，通过保留子空间构建、正交引导及重叠自适应校准，在多个基准上实现擦除与保留的更好平衡。
AI 中文摘要
概念擦除旨在从预训练的文本到图像扩散模型中移除目标概念，例如受版权保护的风格、可识别的角色或不安全内容，同时保持其生成其他内容的能力。现有的激活引导方法主要基于目标概念构建擦除方向，并在推理时沿该方向调整模型激活。然而，目标概念与保留概念在模型的表示空间中往往存在重叠，因此该方向也包含保留概念所依赖的共享成分。直接沿此方向进行引导可能会抑制保留概念，损害非目标内容的生成。为解决这一问题，我们提出了保留感知激活引导（RASteer），一种无需训练的方法。RASteer首先从要保留的概念中构建保留子空间。随后，保留正交引导（ROS）从擦除方向中移除与该子空间对齐的成分，使引导更针对目标。由于完全移除共享成分可能削弱擦除效果，我们进一步引入了重叠自适应校准（OAC）。在每个层和去噪步骤中，OAC利用擦除方向与保留子空间之间的重叠程度来控制每个共享成分被移除的量，从而在目标擦除与概念保留之间取得平衡。在多个骨干网络和基准上针对不安全内容、实例和艺术风格擦除的实验表明，RASteer在擦除与保留之间实现了更好的平衡，达到或超越了所评估的激活引导和权重编辑基线方法。
英文摘要
Concept erasure aims to remove a target concept, such as a copyrighted style, a recognizable character, or unsafe content, from a pretrained text-to-image diffusion model while preserving its ability to generate other content. Existing activation steering methods build an erasure direction mainly from the target concept and adjust model activations along it at inference time. However, target and retained concepts often overlap in the model's representation space, so this direction also contains shared components that retained concepts rely on. Steering directly along this direction can therefore suppress retained concepts and harm the generation of non-target content. To address this issue, we propose Retain-aware Activation Steering (RASteer), a training-free method. RASteer first builds a retain subspace from the concepts to preserve. Retain-Orthogonal Steering (ROS) then removes components aligned with this subspace from the erasure direction, making steering more specific to the target. Since fully removing the shared components can weaken erasure, we further introduce Overlap-Adaptive Calibration (OAC). At each layer and denoising step, OAC uses the overlap between the erasure direction and the retain subspace to control how much of each shared component is removed, balancing target erasure and concept preservation. Experiments on unsafe-content, instance, and artistic-style erasure across multiple backbones and benchmarks show that RASteer matches or outperforms the activation steering and weight editing baselines we evaluate, achieving a better balance between erasure and preservation.
Comments20 pages. Project page: https://rasteer.cvmlgroup.web.illinois.edu/
