LeMario project trains a JEPA world model using Super Mario Bros gameplay

The LeMario project explores AI modeling with video games. It trains a JEPA world model on Super Mario Bros. The approach uses gameplay data as input. Researchers aim to capture game dynamics in the model. The project highlights the potential of unsupervised learning. Results illustrate how the model predicts game states. The work contributes to research on world modeling. Future extensions may apply the technique to other environments.