Learning 3D Persistent Embodied World Models

NeurIPS 2025

Siyuan Zhou Yilun Du, Yuncong Yang, Lei Han, Peihao Chen, Dit-Yan Yeung, Chuang Gan

Installation

conda create -n xxx python=3.10
conda activate xxx
pip install -r requirements.txt

Data Preparation

Install habitat environment following habitat-sim and habitat-lab.

python data/nav.py

Training

# first training stage
bash scripts/habitat_camera.sh
# second training stage
bash scripts/habitat_map.sh

Citation

If you find our work useful, please consider citing:

@article{zhou2025learning,
  title={Learning 3D Persistent Embodied World Models},
  author={Zhou, Siyuan and Du, Yilun and Yang, Yuncong and Han, Lei and Chen, Peihao and Yeung, Dit-Yan and Gan, Chuang},
  journal={arXiv preprint arXiv:2505.05495},
  year={2025}
}