RESEARCH PAPERYear 2025

Mean Flows for One-step Generative Modeling

Zhengyang Geng; Mingyang Deng; Xingjian Bai; Zico Kolter; Kaiming He

Classification

Major category: Foundational work
Quadrant: Not applicable
Architecture: Not applicable
Prediction paradigm: Not applicable
Subcategories: Diffusion & flow-matching foundations
Source review status: Verified from primary sources

Category review. MeanFlow develops a general average-velocity learning identity and one-step generative objective, tested in image generation. It supplies pre-2026 generative-method background rather than a pretrained language/vision/world module or action interface. Reading evidence

AT A GLANCE

Contribution

A contribution summary has not been added yet.

Abstract

An abstract has not been added yet.

Affiliations

Not listed in the collection.

BibTeX

@inproceedings{mean_flow,
 author = {Geng, Zhengyang and Deng, Mingyang and Bai, Xingjian and Kolter, Zico and He, Kaiming},
 booktitle = {NeurIPS},
 pages = {75460--75482},
 title = {Mean Flows for One-step Generative Modeling},
 volume = {38},
 year = {2025}
}

Classification

Contribution

Abstract

Affiliations

BibTeX

More in this category

Critique of World Model

Precise and Dexterous Robotic Manipulation via Human-in-the-Loop Reinforcement Learning

AdaWorld: Learning Adaptable World Models with Latent Actions