不采样不训练，直接从模型参数中提取图像编辑方向
Self-Attention Decomposition For Training Free Diffusion Editing
- 通过分析自注意力权重矩阵的特征向量，直接获取可解释的编辑方向
- 在多个数据集上实现高质量编辑，编辑速度比基准快60%
- 适合需要快速、无额外成本图像编辑的研究者与开发者
扩散模型在图像生成中表现出极高的保真度，但对其输出进行精确控制以实现目标编辑仍具挑战。实现可控性的关键步骤是识别模型潜在表示中对应语义属性的可解释方向。现有方法通常依赖大量图像采样或训练辅助网络，效率受限。本文提出一种分析性方法，直接从预训练扩散模型的参数中推导出语义编辑方向，无需额外数据或微调。核心洞察是：自注意力权重矩阵编码了训练过程中学习到的数据分布结构信息。通过计算这些权重矩阵的特征向量，我们获得鲁棒且可解释的编辑方向。实验表明，该方法在多个数据集上均能实现高质量编辑，同时将编辑时间相比当前基准显著减少60%。
原文摘要 · Abstract (English)
Diffusion models achieve remarkable fidelity in image synthesis, yet precise control over their outputs for targeted editing remains challenging. A key step toward controllability is to identify interpretable directions in the model's latent representations that correspond to semantic attributes. Existing approaches for finding interpretable directions typically rely on sampling large sets of images or training auxiliary networks, which limits efficiency. We propose an analytical method that derives semantic editing directions directly from the pretrained parameters of diffusion models, requiring neither additional data nor fine-tuning. Our insight is that self-attention weight matrices encode rich structural information about the data distribution learned during training. By computing the eigenvectors of these weight matrices, we obtain robust and interpretable editing directions. Experiments demonstrate that our method produces high-quality edits across multiple datasets while reducing editing time significantly by 60% over current benchmarks.
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。