厘清扩散概率模型的 SNR-t 偏差
Elucidating the SNR-t Bias of Diffusion Probabilistic Models
指出推理阶段去噪样本的 SNR 与时间步错配会导致误差累积，并给出按频率分量分别做差分校正的免训练修正手段，可直接插入现有扩散采样器（DDIM、EDM、FLUX 等）以提升生成质量。
原文依据 this correspondence is disrupted during inference, leading to error accumulation and impairing the generation quality
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论扩散概率模型在广泛的生成任务中展现出卓越性能。然而，我们观察到这类模型常常存在信噪比-时间步（SNR-t）偏差。该偏差指的是在推理阶段，去噪样本的信噪比与其对应时间步之间出现错配。具体而言，训练时样本的信噪比与时间步严格耦合；但在推理时这种对应关系被破坏，从而导致误差累积并损害生成质量。我们提供了全面的经验证据与理论分析来证实这一现象，并提出一种简单而有效的差分校正方法来缓解 SNR-t 偏差。考虑到扩散模型在逆向去噪过程中通常先重建低频成分、再关注高频细节，我们将样本分解为不同频率成分，并对每个成分分别施加差分校正。大量实验表明，我们的方法在多种分辨率的数据集上显著提升了多种扩散模型（IDDPM、ADM、DDIM、A-DPM、EA-DPM、EDM、PFGM++ 与 FLUX）的生成质量，且计算开销可忽略不计。
英文原文摘要 1303 字符
Diffusion Probabilistic Models have demonstrated remarkable performance across a wide range of generative tasks. However, we have observed that these models often suffer from a Signal-to-Noise Ratio-timestep (SNR-t) bias. This bias refers to the misalignment between the SNR of the denoising sample and its corresponding timestep during the inference phase. Specifically, during training, the SNR of a sample is strictly coupled with its timestep. However, this correspondence is disrupted during inference, leading to error accumulation and impairing the generation quality. We provide comprehensive empirical evidence and theoretical analysis to substantiate this phenomenon and propose a simple yet effective differential correction method to mitigate the SNR-t bias. Recognizing that diffusion models typically reconstruct low-frequency components before focusing on high-frequency details during the reverse denoising process, we decompose samples into various frequency components and apply differential correction to each component individually. Extensive experiments show that our approach significantly improves the generation quality of various diffusion models (IDDPM, ADM, DDIM, A-DPM, EA-DPM, EDM, PFGM++, and FLUX) on datasets of various resolutions with negligible computational overhead.