提出稳定一致性调优方法，显著提升生成质量与速度。
Stable Consistency Tuning: Understanding and Improving Consistency Models
- 将扩散模型去噪建模为马尔可夫决策过程，用时序差分学习理解一致性模型训练。
- 在ImageNet-64上实现1步FID 2.42、2步FID 1.55，刷新一致性模型新纪录。
- 适合关注生成模型高效采样与训练机制的科研人员阅读。
扩散模型虽生成质量高，但因迭代去噪导致生成速度慢。一致性模型作为新一代生成模型，在显著加快采样速度的同时，仍能保持良好性能。其训练方式包括利用预训练扩散模型的一致性蒸馏，或直接从原始数据进行一致性训练/调优。本文通过将扩散模型的去噪过程建模为马尔可夫决策过程（MDP），并把一致性模型训练视为时序差分（TD）学习中的价值估计，提出了新的理论框架。该框架揭示了现有训练策略的局限性。在此基础上，基于易用一致性调优（ECT），我们提出稳定一致性调优（SCT），引入基于得分恒等式的方差减少学习。SCT在CIFAR-10和ImageNet-64等基准上表现显著提升，在ImageNet-64上实现1步FID 2.42、2步FID 1.55，成为一致性模型的新最优结果。
原文摘要 · Abstract (English)
Diffusion models achieve superior generation quality but suffer from slow generation speed due to the iterative nature of denoising. In contrast, consistency models, a new generative family, achieve competitive performance with significantly faster sampling. These models are trained either through consistency distillation, which leverages pretrained diffusion models, or consistency training/tuning directly from raw data. In this work, we propose a novel framework for understanding consistency models by modeling the denoising process of the diffusion model as a Markov Decision Process (MDP) and framing consistency model training as the value estimation through Temporal Difference~(TD) Learning. More importantly, this framework allows us to analyze the limitations of current consistency training/tuning strategies. Built upon Easy Consistency Tuning (ECT), we propose Stable Consistency Tuning (SCT), which incorporates variance-reduced learning using the score identity. SCT leads to significant performance improvements on benchmarks such as CIFAR-10 and ImageNet-64. On ImageNet-64, SCT achieves 1-step FID 2.42 and 2-step FID 1.55, a new SoTA for consistency models.
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。