通过一对一映射实现扩散模型中概念的精准擦除
SAEmnesia: Erasing Concepts in Diffusion Models with Supervised Sparse Autoencoders
- 训练时标注概念，使每个概念绑定唯一可解释神经元
- 比现有方法减少96.67%超参搜索，对象擦除准确率提升9.22%
- 支持连续擦除多个概念，且对对抗攻击保持鲁棒
扩散模型中的概念去学习受限于特征分裂问题，即概念分散在多个潜在特征中，导致移除困难且计算成本高。本文提出SAEmnesia，一种基于监督稀疏自编码器的框架，通过强制实现概念与神经元的一一对应关系来克服该问题。训练过程中系统性标注概念，实现特征集中化，使每个概念仅由一个可解释神经元承载，从而实现高度精准和高效的的概念擦除。相比当前最先进的基于稀疏自编码器的去学习方法，SAEmnesia将超参数搜索量降低96.67%，在UnlearnCanvas基准上对物体的去学习表现提升9.22%。该方法在连续去学习场景中也展现出优异可扩展性，当移除九个物体时准确率提升28.4%，显著推进了精确可控的概念擦除。此外，SAEmnesia在I2P基准上有效抑制了裸露内容，并对对抗攻击保持稳健。源代码已开源：https://github.com/EIDOSLAB/SAEmnesia。
原文摘要 · Abstract (English)
Concept unlearning in diffusion models is hampered by feature splitting, where concepts are distributed across many latent features, making their removal challenging and computationally expensive. We introduce SAEmnesia, a supervised sparse autoencoder framework that overcomes this by enforcing one-to-one concept-neuron mappings. By systematically labeling concepts during training, our method achieves feature centralization, binding each concept to a single, interpretable neuron. This enables highly targeted and efficient concept erasure. Compared to the state-of-the-art sparse autoencoder-based unlearning approach, SAEmnesia reduces hyperparameter search by 96.67% and achieves a 9.22% improvement on the UnlearnCanvas benchmark for objects. Our method also shows superior scalability in sequential unlearning, improving accuracy by 28.4% when removing nine objects, establishing a step forward for precise and controllable concept erasure. Moreover, SAEmnesia effectively suppresses nudity on the I2P benchmark and remains robust to adversarial attacks. Source code available at https://github.com/EIDOSLAB/SAEmnesia.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。