Diffusion Probabilistic Models (DPM) have shown remarkable efficacy in the synthesis of high-quality images. However, their inference process characteristically requires numerous, potentially hundreds, of iterative steps, which could exaggerate the problem of exposure bias due to the training and inference discrepancy. Previous work has attempted to mitigate this issue by perturbing inputs during training, which consequently mandates the retraining of the DPM. In this work, we conduct a systematic study of exposure bias in DPM and, intriguingly, we find that the exposure bias could be alleviated with a novel sampling method that we propose, without retraining the model. We empirically and theoretically show that, during inference, for each backward time step $t$ and corresponding state $\hat{x}_t$, there might exist another time step $t_s$ which exhibits superior coupling with $\hat{x}_t$. Based on this finding, we introduce a sampling method named Time-Shift Sampler. Our framework can be seamlessly integrated to existing sampling algorithms, such as DDPM, DDIM and other high-order solvers, inducing merely minimal additional computations. Experimental results show our method brings significant and consistent improvements in FID scores on different datasets and sampling methods. For example, integrating Time-Shift Sampler to F-PNDM yields a FID=3.88, achieving 44.49\% improvements as compared to F-PNDM, on CIFAR-10 with 10 sampling steps, which is more performant than the vanilla DDIM with 100 sampling steps. Our code is available at https://github.com/Mingxiao-Li/TS-DPM.

翻译：扩散概率模型在高质量图像合成中展现出显著效能。然而其推理过程通常需要数百次迭代步骤，这可能导致训练与推理之间的差异加剧暴露偏差问题。先前的工作试图通过在训练过程中扰动输入来缓解该问题，但这要求重新训练扩散模型。在本工作中，我们系统研究了扩散模型中的暴露偏差，并有趣地发现，通过我们提出的新型采样方法，无需重新训练模型即可缓解暴露偏差。我们通过理论与实验证明，在推理过程中，对于每个反向时间步 $t$ 及其对应状态 $\hat{x}_t$，可能存在另一个时间步 $t_s$ 与 $\hat{x}_t$ 具有更优的耦合性。基于这一发现，我们提出名为时间偏移采样器的采样方法。该框架可无缝集成至现有采样算法（如DDPM、DDIM及其他高阶求解器），仅引入极少量额外计算。实验结果表明，我们的方法在不同数据集和采样方法上均能显著且一致地提升弗雷歇初始距离得分。例如，在CIFAR-10数据集上使用10步采样时，将时间偏移采样器集成至F-PNDM可获得FID=3.88的得分，较F-PNDM提升44.49%，优于使用100步采样的原始DDIM。我们的代码可在https://github.com/Mingxiao-Li/TS-DPM 获取。