zhefeigong/korr

Robust Online Residual Refinement
via Koopman-Guided Dynamics Modeling

Zhefei Gong¹, Shangke Lyu^1✉, Pengxiang Ding¹², Wei Xiao¹, Donglin Wang^1✉

¹Westlake University, ²Zhejiang University

👀 Overview

TL;DR: introduce KORR (Koopman-guided Online Residual Refinement), a simple yet effective framework that conditions residual corrections on Koopman-predicted latent states, enabling globally informed and stable action refinement.

🙏 Acknowledgment

We sincerely thank the authors of Furniture-Bench for providing a high-quality benchmark environment, and appreciate the insightful preliminary exploration of residual policy learning in From Imitation to Refinement, which inspired part of our work.

🏷️ License

This repository is licensed under the MIT License. See the LICENSE file for more details.

📌 Citation

If you find our work useful, please consider citing the following paper:

@misc{gong2025robustonlineresidualrefinement,
      title={Robust Online Residual Refinement via Koopman-Guided Dynamics Modeling}, 
      author={Zhefei Gong and Shangke Lyu and Pengxiang Ding and Wei Xiao and Donglin Wang},
      year={2025},
      eprint={2509.12562},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2509.12562}, 
}