Building Flow Uniqueness in One-step Generative Modeling

Junyu Zhang, Daochang Liu, Liu.Liu, Zhizhong Su, Jong Hwan Ko, Shichao Zhang, Eunbyung Park, Chang Xu

Keywords: Flow matching, one-step generation, flow consistency

Abstract:

Recent advances in generative modeling frameworks, such as diffusion models and flow matching, have achieved record-breaking performance.
Nevertheless, these approaches involve iterative sampling procedures across many neural network passes, which severely limits their practical deployment, particularly in domains demanding real-time interaction.
Although considerable effort has been devoted to accelerating sampling, achieving high-quality one-step generation remains an open challenge, motivating research into a new era of generative modeling.
Motivated by this, we put forward a novel and effective framework, termed \textit{Flow Uniqueness Models} (\textbf{FUM}).
The core idea of FUM is to construct strictly one-to-one image pairs, thereby enforcing velocity uniqueness along the entire sampling path, which forms as the foundation for few-step sampling.
By leveraging this modeling mechanism, FUM not only achieves remarkable one-step generative performance but also provides the flexibility to balance image quality against the number of sampling steps.
Extensive experiments on three benchmark datasets comprehensively validate the superiority of our proposed FUM.

Primary Area: generative models

Submission Number: 6913