Coefficients-Preserving Sampling for Reinforcement Learning with Flow Matching

paper link: https://arxiv.org/abs/2509.05952

中文介绍: https://zhuanlan.zhihu.com/p/1948388095151026330

Changelog

2025/09/09: Please check this commit for the implementation of Flow-CPS. We also provide the code to generate Figure 2 in this notebook

Installation and Usage

Please follow the instructions in: https://github.com/yifan123/flow_grpo

🤗 Acknowledgement

This repo is based on flow-grpo, ddpo-pytorch and diffusers. We thank the authors for their valuable contributions to the AIGC community. Special thanks to Kevin Black for the excellent ddpo-pytorch repo.

⭐Citation

If you find Flow-CPS useful for your research or projects, we would greatly appreciate it if you could cite the following paper:

@article{wang2025coefficients,
  title={Coefficients-Preserving Sampling for Reinforcement Learning with Flow Matching},
  author={Wang, Feng and Yu, Zihao},
  journal={arXiv preprint arXiv:2509.05952},
  year={2025}
}