We propose an evolution-based approach to meta-learn synthetic neural environments and reward neural networks for reinforcement learning. - View it on GitHub