Version 1.0 arXiv July 2025 Release

The code in this release should be able to recreate results from our recent arXiv paper:
https://arxiv.org/abs/2507.00184
You can use the code to checkout pre-trained diffusion models from Hugging Face, or you can train your own models. Please use our interactive GUI to generate Mario scenes and combine them into complete levels!

If you notice any problems or have any questions about the code, feel free to open a new Issue or contact Dr. Jacob Schrum:
[email protected]