arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2510.21792cs.LGcs.AI
收藏

方差缩减引导:扩散模型的采样轨迹优化

Variance-Reduction Guidance: Sampling Trajectory Optimization for Diffusion Models

College of Computing;Data Science, Nanyang Technological University Singapore;Institute for Infocomm Research (I2R) \& Centre for Frontier AI Research, A STAR Singapore liu\

Shifeng Xu, Yanzhu Liu, Adams Wai-Kin Kong

AI总结:

针对扩散模型采样中预测误差累积导致生成质量下降的问题,提出方差缩减引导(VRG)方法,无需微调即可优化采样轨迹,显著提升生成质量。

摘要:

扩散模型已成为新兴的生成模型。其采样过程涉及多个步骤,每一步中模型从带噪样本预测噪声。当模型进行预测时,输出会偏离真实值,我们将这种偏差称为“预测误差”。预测误差在采样过程中累积,并降低生成质量。本文介绍了一种统计测量预测误差的新技术,并提出了方差缩减引导(VRG)方法来减轻该误差。VRG无需模型微调或修改。给定预定义的采样轨迹,它搜索一条具有相同采样步数但产生更高质量结果的新轨迹。VRG适用于条件生成和无条件生成。在多种数据集和基线上的实验表明,VRG能显著提高扩散模型的生成质量。源代码可在https://github.com/shifengxu/VRG获取。

英文摘要:

Diffusion models have become emerging generative models. Their sampling process involves multiple steps, and in each step the models predict the noise from a noisy sample. When the models make prediction, the output deviates from the ground truth, and we call such a deviation as \textit{prediction error}. The prediction error accumulates over the sampling process and deteriorates generation quality. This paper introduces a novel technique for statistically measuring the prediction error and proposes the Variance-Reduction Guidance (VRG) method to mitigate this error. VRG does not require model fine-tuning or modification. Given a predefined sampling trajectory, it searches for a new trajectory which has the same number of sampling steps but produces higher quality results. VRG is applicable to both conditional and unconditional generation. Experiments on various datasets and baselines demonstrate that VRG can significantly improve the generation quality of diffusion models. Source code is available at https://github.com/shifengxu/VRG.

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。

↑