Research stack for reproducible federated training and evaluation of action-conditioned JEPA world models. - View it on GitHub