LAWAM LIBERO SFT Release

This repository contains the LAWAM SFT checkpoint for LIBERO experiments, presented in the paper LaWAM: Latent World Action Models for Efficient Dynamics-Aware Robot Policies.

• Project Page: https://rlinf.github.io/LaWAM/
• Repository: https://github.com/RLinf/LaWAM

Dataset Notes

The LIBERO training data used for this checkpoint is derived from IPEC-COMMUNITY/libero-benchmark-dataset.

Compared with the public LIBERO benchmark datasets, our processed version has the following differences:

• We merge the four LIBERO subsets into a unified training mixture.
• We convert the merged data into LeRobot 3.0 format.
• The accompanying dataset_statistics.json and config.yaml correspond to this processed LeRobot 3.0 version, not the raw public release.

Files

• final_model/pytorch_model.pt: released LAWAM checkpoint.
• config.yaml: training/configuration metadata for this release.
• dataset_statistics.json: dataset statistics for the processed training data.

Citation

@misc{chen2026lawam,
  title = {LaWAM: Latent World Action Models for Efficient Dynamics-Aware Robot Policies},
  author = {Chen, Jialei and Wang, Kai and Chen, Kang and Chen, Shuaihang and Gao, Feng and Tang, Wenhao and Li, Zhiyuan Ext, Weilin and Yao, Zhuyu and Li, Boxun and Yuanbo, Xu and Yu, Chao},
  journal = {arXiv preprint arXiv:2606.15768},
  year = {2026},
  archiveprefix = {arXiv},
  primaryclass = {cs.RO},
}