Mean flow (MeanFlow) enables efficient, high-fidelity image generation, yet its single-function evaluation (1-NFE) generation often cannot yield compelling results. We address this issue by introducing RMFlow, an efficient multimodal generative model that integrates a coarse 1-NFE MeanFlow transport with a subsequent tailored noise-injection refinement step. RMFlow approximates the average velocity of the flow path using a neural network trained with a new loss function that balances minimizing the Wasserstein distance between probability paths and maximizing sample likelihood. RMFlow achieves near state-of-the-art results on text-to-image, context-to-molecule, and time-series generation using only 1-NFE, at a computational cost comparable to the baseline MeanFlows.
@article{arxiv.2602.00849,
title = {RMFlow: Refined Mean Flow by a Noise-Injection Step for Multimodal Generation},
author = {Yuhao Huang and Shih-Hsin Wang and Andrea L. Bertozzi and Bao Wang},
journal= {arXiv preprint arXiv:2602.00849},
year = {2026}
}