arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.20360cs.LGcs.CV
收藏

动量引导:用于流模型的即插即用引导

Momentum Guidance: Plug-and-Play Guidance for Flow Models

The University of Texas at Austin(德克萨斯大学奥斯汀分校)

Runlong Liao, Jian Yu, Baiyu Su, Chi Zhang, Lizhang Chen, Qiang Liu

AI总结:

本文提出动量引导方法,通过扩展当前速度远离过去速度的指数移动平均,提升样本质量并保持成本,优于现有引导技术,实验显示在ImageNet-256上FID显著降低。

摘要:

流形生成方法提供了一种简单有效的框架用于高保真的生成,但预训练的流模型很少以原始条件形式使用:在图像生成中,没有引导的样本往往显得模糊且缺乏细粒度细节。现有的引导技术如分类器自由引导(CFG)提高了保真度但降低了样本多样性。我们引入了动量引导(MG),一种引导方法,通过沿ODE轨迹扩展当前速度远离过去速度的指数移动平均来提高样本质量,同时保持标准的每步一次评估成本。MG在CFG之外提供了提升,改善了精度-召回率帕累托前沿。实验展示了MG在各种基准上的有效性。在ImageNet-256上,MG在无CFG的情况下提高了36.54%的FID,在有CFG的情况下平均提高了25.42%,在16次采样步骤时达到了1.553的FID。对大型流基模型,包括Stable Diffusion 3和FLUX.1-dev的评估进一步证实了在标准指标上的改进。

英文摘要:

Flow-based generative methods offer a simple and effective framework for high-fidelity generation, yet pretrained flow models are rarely used in their vanilla conditional form: in image generation, samples without guidance often appear diffuse and lack fine-grained detail. Existing guidance techniques such as classifier-free guidance (CFG) improve fidelity but reduce sample diversity. We introduce Momentum Guidance (MG), a guidance method that improves sample quality by extrapolating the current velocity away from an exponential moving average of past velocities along the ODE trajectory, while preserving the standard one-evaluation-per-step cost. MG provides gains beyond CFG, improving the precision-recall Pareto frontier. Experiments demonstrate the effectiveness of MG across benchmarks. On ImageNet-256, MG improves FID by 36.54% without CFG and 25.42% with CFG on average across sampling settings, attaining an FID of 1.553 at 16 sampling steps. Evaluations on large flow-based models, including Stable Diffusion 3 and FLUX.1-dev, further confirm improvements across standard metrics.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑