WARP-RM

A Warp-Augmented Relative Progress Reward Model for Data Curation

Project Page  |  arXiv (coming soon)  

---

WARP-RM is a self-supervised reward model that learns dense relative-progress signals from demonstrations and uses them to reweight action chunks during behavior cloning for long-horizon robot manipulation.

The code used in this project to deploy high-frequency visuomotor policies on the bimanual I2RT YAM robot is hosted on repo uynitsuj/robots_realtime

TODOs

• Coming Soon! Initial code release

Citation

BibTeX coming soon. See the project page in the meantime.

License

This project is released under the MIT License.