[NeurIPS2025] Bootstrap Off-policy with World Model (BOOM)

Quick Start Instructions

conda create -n boom python=3.10
conda activate boom
pip install -e .  # Ensure setup.py exists in the current directory
cd boom
python train.py # Default is dog-run, with an expected score of ~820.
# For Humanoid-bench, you can set task=humanoid_h1hand-run-v0 as an example.
# Note that we use the h1hand version, whose obs dim is 151 and action dim is 61.

Bibtex citation

If you find our work useful, please cite our paper:

@inproceedings{boom2025neurips,
  title={Bootstrap Off-policy with World Model},
  author={Zhan, Guojian and Wang, Likun and Zhang, Xiangteng and Gao, Jiaxin, Tomizuka, Masayoshi and Li, Shengbo Eben},
  booktitle={Advances in Neural Information Processing Systems (NeurIPS)},
  year={2025}
}