COvolve: Adversarial Co-Evolution of Large-Language-Model-Generated Policies and Environments via Two-Player Zero-Sum Game

1. 1Generate environments and agent policies using large language models
2. 2Model the interaction between environments and agents as a two-player zero-sum game
3. 3Use the game framework to co-evolve environments and policies
4. 4Evaluate and refine the co-evolved environments and policies

AI researchers and engineers on a team can benefit from COvolve as it enables continual learning and generalization of agents beyond the training distribution, while also improving the robustness of environments

💡 Co-evolving environments and agent policies can lead to continually improving agents and more robust environments

🤖 COvolve: co-evolving environments & agent policies with LLMs & 2-player zero-sum games