CURE：通过扩散模型中的正交表示编辑实现概念遗忘
CURE: Concept Unlearning via Orthogonal Representation Editing in Diffusion Models
AI总结：
本文提出CURE，一个无需训练的扩散模型概念遗忘框架，通过闭式正交投影模块Spectral Eraser在权重空间中识别并隔离不良概念特征，实现快速、彻底且对无关能力损害极小的内容移除，并增强对红队攻击的鲁棒性。
摘要：
随着文本到图像模型的不断演进，生成不安全、侵犯版权或违反隐私内容的风险也随之增加。现有的安全干预措施——从训练数据筛选、模型微调到推理时过滤和引导——往往存在概念移除不彻底、易受越狱攻击、计算效率低下或对无关能力产生附带损害等问题。在本文中，我们提出了CURE，这是一个无需训练的概念遗忘框架，直接在预训练扩散模型的权重空间中操作，实现对不良概念的快速、可解释且高度特异性的抑制。我们方法的核心是Spectral Eraser，这是一个闭式正交投影模块，通过对与待遗忘和待保留概念相关的token嵌入进行奇异值分解来识别判别性子空间。直观地说，Spectral Eraser识别并隔离不良概念独有的特征，同时保留安全属性。然后，该算子通过单步更新应用，生成一个已编辑的模型，其中目标概念被有效遗忘——无需重新训练、监督或迭代优化。为了平衡过滤毒性与保留无关概念之间的权衡，我们进一步引入了用于谱正则化的Expansion Mechanism，它根据奇异向量与遗忘强度控制的相对重要性，选择性地调制奇异向量。上述所有过程均为闭式，确保仅在$2$秒内实现极其高效的擦除。与先前方法进行基准测试，CURE对目标艺术风格、物体、身份或露骨内容实现了更高效、更彻底的移除，对原始生成能力的损害极小，并展示了对红队测试的增强鲁棒性。
英文摘要：
As Text-to-Image models continue to evolve, so does the risk of generating unsafe, copyrighted, or privacy-violating content. Existing safety interventions - ranging from training data curation and model fine-tuning to inference-time filtering and guidance - often suffer from incomplete concept removal, susceptibility to jail-breaking, computational inefficiency, or collateral damage to unrelated capabilities. In this paper, we introduce CURE, a training-free concept unlearning framework that operates directly in the weight space of pre-trained diffusion models, enabling fast, interpretable, and highly specific suppression of undesired concepts. At the core of our method is the Spectral Eraser, a closed-form, orthogonal projection module that identifies discriminative subspaces using Singular Value Decomposition over token embeddings associated with the concepts to forget and retain. Intuitively, the Spectral Eraser identifies and isolates features unique to the undesired concept while preserving safe attributes. This operator is then applied in a single step update to yield an edited model in which the target concept is effectively unlearned - without retraining, supervision, or iterative optimization. To balance the trade-off between filtering toxicity and preserving unrelated concepts, we further introduce an Expansion Mechanism for spectral regularization which selectively modulates singular vectors based on their relative significance to control the strength of forgetting. All the processes above are in closed-form, guaranteeing extremely efficient erasure in only $2$ seconds. Benchmarking against prior approaches, CURE achieves a more efficient and thorough removal for targeted artistic styles, objects, identities, or explicit content, with minor damage to original generation ability and demonstrates enhanced robustness against red-teaming.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
