Reliable World Simulation for Autonomous Driving

Jiazhi Yang, Kashyap Chitta, Shenyuan Gao, Long Chen, Yuqian Shao, Xiaosong Jia, Hongyang Li, Andreas Geiger, Xiangyu Yue, Li Chen

📜 [technical report], 🎬 [video demos]

📧 Primary Contact: Jiazhi Yang ([email protected])

• ReSim is a driving world model that enables Reliable Simulation of diverse open-world driving scenarios under various actions, including hazardous non-expert ones. A Video2Reward model estimates the reward from ReSim’s simulated future.

• The key ingredient is to co-train the world model on heterogeneous driving data including driving videos from the web, driving data with action labels, and simulated data with non-expert driving behaviors.

• The high-fidelity prediction, accurate action-following, and reward estimation abilities of ReSim facilitate multiple driving applications.

• Code release (Estimated in July).
• Pretrained weights for ReSim world model.
• Simulated data from CARLA with non-expert behaviors.

If any parts of our paper and code help your research, please consider citing us and giving a star to our repository.

@article{yang2025resim,
  title={ReSim: Reliable World Simulation for Autonomous Driving},
  author={Jiazhi Yang and Kashyap Chitta and Shenyuan Gao and Long Chen and Yuqian Shao and Xiaosong Jia and Hongyang Li and Andreas Geiger and Xiangyu Yue and Li Chen},
  journal={arXiv preprint arXiv:2506.09981},
  year={2025}
}