免训练、感知一致的低分辨率预览与高分辨率图像：面向扩散模型高效工作流
Training-free, Perceptually Consistent Low-Resolution Previews with High-Resolution Image for Efficient Workflows of Diffusion Models
对做扩散/流匹配采样加速的人，这篇给出一个免训练的「commutator-zero」条件：通过选择下采样矩阵并施加引导，让低分辨率预览与高分辨率结果在感知上保持一致，可直接作为 HR 生成前的候选筛选环节，与现有少步/蒸馏加速叠加还能再拿 3 倍加速。
原文依据 We propose the commutator-zero condition to ensure the LR-HR perceptual consistency for flow matching models
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论图像生成模型已成为从普通用户到专业设计师人人可用的、产出精美高分辨率（HR）图像不可或缺的工具。然而，要得到理想结果，往往需要用不同的提示词和随机种子生成大量 HR 图像，这给用户和服务提供商都带来高昂的计算开销。先生成低分辨率（LR）图像可以缓解计算负担，但如何生成与其 HR 对应结果在感知上一致的低分辨率图像并不简单。本文研究生成高保真 LR 图像（称为 Previews）的任务，这些预览需保持与其 HR 对应结果的感知相似性，以支撑高效工作流，使用户能在生成最终 HR 图像之前识别出有希望的候选。我们提出 commutator-zero 条件，为流匹配模型保证 LR-HR 感知一致性，由此得到一种免训练方案，包含下采样矩阵选择与 commutator-zero 引导。大量实验表明，我们的方法可在保持 HR 感知一致性的同时生成 LR 图像，计算量最多降低 33%。与现有加速技术结合时，我们的方法可实现最高 3 倍加速。此外，我们的形式化表述可扩展到图像变换操作，例如形变与平移，展示了其泛化性。
英文原文摘要 1411 字符
Image generative models have become indispensable tools to yield exquisite high-resolution (HR) images for everyone, ranging from general users to professional designers. However, a desired outcome often requires generating a large number of HR images with different prompts and seeds, resulting in high computational cost for both users and service providers. Generating low-resolution (LR) images first could alleviate computational burden, but it is not straightforward how to generate LR images that are perceptually consistent with their HR counterparts. Here, we consider the task of generating high-fidelity LR images, called Previews, that preserve perceptual similarity of their HR counterparts for an efficient workflow, allowing users to identify promising candidates before generating the final HR image. We propose the commutator-zero condition to ensure the LR-HR perceptual consistency for flow matching models, leading to the proposed training-free solution with downsampling matrix selection and commutator-zero guidance. Extensive experiments show that our method can generate LR images with up to 33% computation reduction while maintaining HR perceptual consistency. When combined with existing acceleration techniques, our method achieves up to 3xspeedup. Moreover, our formulation can be extended to image manipulations, such as warping and translation, demonstrating its generalizability.