发现扩散模型训练中表示退化问题，提出动态优化框架提升收敛速度与生成质量。
Elucidating Representation Degradation Problem in Diffusion Model Training
- 基于有效可恢复性动态重分配优化资源，缓解噪声下的结构失真
- 在多个扩散模型上实现更快收敛，生成质量显著提升
- 无需额外监督，适配主流扩散模型架构，提升训练效率
扩散模型虽取得显著成功，但其训练效率受限于严重的优化瓶颈，我们称之为表示退化。随着噪声水平升高，模型输出出现渐进式结构失真，导致训练不稳定并降低生成质量。分析表明，这种不稳定性源于目标可恢复性不匹配，与神经正切核（NTK）谱衰减及有效低秩行为相关。为此，我们提出阐明表示扩散（ERD），一种即插即用的框架，根据有效可恢复性动态重分配优化努力。通过在无外部监督下稳定表示学习，ERD加速收敛，并在多种扩散主干网络上实现优异的实证性能。
原文摘要 · Abstract (English)
Diffusion models have achieved remarkable success, yet their training remains inefficient due to a severe optimization bottleneck, which we term Representation Degradation. As noise levels increase, the outputs of the trained model exhibit progressive structural distortion, which can destabilize training and impair generation quality. Our analysis suggests that this instability is driven by mismatched target recoverability, which is associated with Neural Tangent Kernel (NTK) spectral weakening and effective low-rank behavior. To address this, we propose Elucidated Representation Diffusion (ERD), a plug-and-play framework that dynamically reallocates optimization effort according to effective recoverability. By stabilizing representation learning without external supervision, ERD accelerates convergence and achieves strong empirical performance across diffusion backbones.
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。