arXiv：cs.LG· Juyeop Han, Lukas Lao Beyer, Sertac Karaman·· 7 小时前
UA-Flow：为流匹配引入不确定性量化与引导
Flow Matching with Uncertainty Quantification and Guidance
AI 导读
研究者提出不确定性感知流匹配 UA-Flow，在预测速度场的同时预测异方差不确定性，并通过流动力学传播得到空间局部化的样本不确定性。
正文
Abstract:Despite the remarkable success of sampling-based generative models such as flow matching, they can still produce samples of inconsistent or degraded quality. To assess sample reliability and improve sample fidelity, we propose uncertainty-aware flow matching (UA-Flow), a lightweight extension of flow matching that predicts the velocity field together with heteroscedastic uncertainty. UA-Flow estimates spatially localized sample uncertainty by propagating velocity uncertainty through the flow dynamics. The resulting estimates serve as reliability signals that reflect data-manifold support for each generated sample or region. We further use them to steer generation via uncertainty-aware guidance (UG) and uncertainty-aware classifier-free guidance (UCFG). Experiments on ImageNet and CIFAR-10 show that UA-Flow's uncertainty tracks sample fidelity more consistently than element-wise baselines and that these signals also carry information about data-manifold support. On ImageNet, UG increases precision at the cost of recall and improves FID, while UCFG maintains lower FID than CFG at large guidance scales.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG) |
| Cite as: | arXiv:2602.10326 [cs.CV] |
| (or arXiv:2602.10326v3 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2602.10326 arXiv-issued DOI via DataCite |
Submission history
From: Juyeop Han [view email]
[v1]
Tue, 10 Feb 2026 22:03:13 UTC (36,577 KB)
[v2]
Tue, 12 May 2026 20:51:48 UTC (38,409 KB)
[v3]
Wed, 14 Oct 2026 01:30:26 UTC (40,550 KB)
来源：arXiv：cs.LG · arxiv.org