Continual Bench

Continual Bench provides an environment suitable for evaluating online reinforcement learning agents under the continual learning setup, with a unified world dynamics.

git clone git@github.com:sail-sg/ContinualBench.git && cd ContinualBench
pip install -e .

from continual_bench.envs import ContinualBenchEnv

env = ContinualBenchEnv(render_mode="rgb_array", seed=0)
action = ...
next_obs, reward, terminated, truncated, info = env.step(action)

continual-bench is distributed under the terms of the MIT license.

If you find our work useful for your research, please consider citing:

@inproceedings{liu2025continual,
  title={Continual Reinforcement Learning by Planning with Online World Models},
  author={Liu, Zichen and Fu, Guoji and Du, Chao and Lee, Wee Sun and Lin, Min},
  booktitle={International Conference on Machine Learning},
  year={2025}
}