通过概念域校正与概念保持梯度消除扩散模型中的概念
Unlearning Concepts in Diffusion Model via Concept Domain Correction and Concept Preserving Gradient
AI总结：
针对扩散模型机器遗忘中泛化不足与效用退化问题，提出 DoCo 概念域校正框架及概念保持梯度手术，在多种概念和分布外提示上实现更有效且低损伤的概念遗忘。
摘要：
文本到图像扩散模型在生成照片级真实图像方面取得了显著成功。然而，预训练过程中包含敏感信息会带来重大风险。机器遗忘（Machine Unlearning，MU）为从这些模型中消除敏感概念提供了一种有前景的解决方案。尽管具有潜力，现有 MU 方法仍面临两大挑战：1）泛化能力有限，概念擦除仅在已遗忘集合内有效，无法阻止集合外提示生成敏感概念；2）效用退化，移除目标概念会显著影响模型的整体性能。为解决这些问题，我们提出了一个名为 DoCo（Domain Correction，域校正）的新型概念域校正框架。通过对抗训练对齐敏感概念与锚点概念的输出域，我们的方法可确保对目标概念进行全面遗忘。此外，我们引入了一种概念保持梯度手术技术，用于缓解相互冲突的梯度分量，从而在遗忘特定概念的同时保持模型效用。针对各种实例、风格和冒犯性概念进行的大量实验表明，我们的方法能够有效遗忘目标概念，并对相关概念影响极小；即使对于分布外提示，其表现也优于以往方法。
英文摘要：
Text-to-image diffusion models have achieved remarkable success in generating photorealistic images. However, the inclusion of sensitive information during pre-training poses significant risks. Machine Unlearning (MU) offers a promising solution to eliminate sensitive concepts from these models. Despite its potential, existing MU methods face two main challenges: 1) limited generalization, where concept erasure is effective only within the unlearned set, failing to prevent sensitive concept generation from out-of-set prompts; and 2) utility degradation, where removing target concepts significantly impacts the model's overall performance. To address these issues, we propose a novel concept domain correction framework named \textbf{DoCo} (\textbf{Do}main \textbf{Co}rrection). By aligning the output domains of sensitive and anchor concepts through adversarial training, our approach ensures comprehensive unlearning of target concepts. Additionally, we introduce a concept-preserving gradient surgery technique that mitigates conflicting gradient components, thereby preserving the model's utility while unlearning specific concepts. Extensive experiments across various instances, styles, and offensive concepts demonstrate the effectiveness of our method in unlearning targeted concepts with minimal impact on related concepts, outperforming previous approaches even for out-of-distribution prompts.
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。
