150 games. Keyboard required.
Every game is also an RL environment you can train an agent on
from the same file. Train here.
Craftax: M. Matthews, M. Beukman, B. Ellis, M. Samvelyan, M. Jackson, S. Coward and J. Foerster, Craftax: A Lightning-Fast Benchmark for Open-Ended Reinforcement Learning, ICML 2024. Dynamics matched against the C implementation in PufferLib (J. Suarez, 2024).
Levels from DMLab-30: C. Beattie, J. Z. Leibo, D. Teplyashin, T. Ward, M. Wainwright, H. Küttler, A. Lefrancq, S. Green, V. Valdés, A. Sadik et al., DeepMind Lab, arXiv 2016. Layouts, movement and rewards checked frame by frame against DeepMind Lab itself. Textures from its //assets, CC BY 4.0.
JavaScript clones of the ProcGen games: K. Cobbe, C. Hesse, J. Hilton and J. Schulman, Leveraging Procedural Generation to Benchmark Reinforcement Learning, ICML 2020.
JavaScript clones of Atari 2600 games from the Arcade Learning Environment: M. G. Bellemare, Y. Naddaf, J. Veness and M. Bowling, The Arcade Learning Environment: An Evaluation Platform for General Agents, JAIR 2013.
Games and reference emulator from Octax: W. Radji, T. Michel and H. Piteau, Octax: Accelerated CHIP-8 Arcade Environments for Reinforcement Learning in JAX, ICLR 2026.
VGDL: T. Schaul, A Video Game Description Language for Model-based or Interactive Learning, IEEE CIG 2013. Engine checked against py-vgdl as ported by M. Tomov.
The games from M. S. Tomov, P. A. Tsividis, T. Pouncy, J. B. Tenenbaum and S. J. Gershman, The neural architecture of theory-based reinforcement learning, Neuron 2023. Engine checked against their RC_RL code. These are the study's v4 versions.
PuzzleScript by Stephen Lavelle (increpare); engine and games from github.com/increpare/PuzzleScript. Engine © 2013 Stephen Lavelle, MIT License. Each game is by the author shown under it.
Games by other authors are credited under their names and on their pages. If one of them is yours and you would like it removed, email [email protected] and it will be taken down.