Utkarsh* • Pengfei Cai* • Alan Edelman • Rafael Gómez-Bombarelli • Christopher Rackauckas
To appear at NeurIPS 2025.
This repo implements Physics-Constrained Flow Matching (PCFM) -- a framework that enforces physical constraints during sampling of flow-based generative models.
PCFM is released under the MIT License.
This repository includes components derived from amazon-science/ECI-sampling licensed under the Apache License 2.0. See LICENSE, LICENSE-APACHE-2.0, and NOTICE for details.
If you use this repository, please cite:
@article{PCFM2025,
title={Physics-Constrained Flow Matching: Sampling Generative Models with Hard Constraints},
author={Utkarsh, Utkarsh and Cai, Pengfei and Edelman, Alan and Gomez-Bombarelli, Rafael and Rackauckas, Christopher Vincent},
journal={arXiv preprint arXiv:2506.04171},
year={2025}
}