Sylvest/pick-place-wcm-ckpt

WCM: A World Critic Model for Vision-Language-Action Reinforcement Learning

This repository contains model artifacts for WCM (World Critic Model), presented in WCM: A World Critic Model for Vision-Language-Action Reinforcement Learning.

• Project Page: https://sylvestf.github.io/wcm-homepage/
• GitHub Repository: https://github.com/sylvestf/WCM
• Paper: https://huggingface.co/papers/2607.29613

Overview

WCM is a history-aware critic for partially observable robot control. Built on a lightweight LeJEPA architecture, WCM jointly learns to estimate the value of the current state and predict the next latent state, equipping Vision-Language-Action (VLA) reinforcement learning with a representation explicitly trained to capture temporal dynamics rather than merely regressing scalar returns.

Citation

@misc{fei2026wcmworldcriticmodel,
      title={WCM: A World Critic Model for Vision-Language-Action Reinforcement Learning}, 
      author={Senyu Fei and Xiaopeng Yu and Siyin Wang and Xianzhong Zhao and Jingjing Gong and Xipeng Qiu},
      year={2026},
      eprint={2607.29613},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2607.29613}, 
}