Tag
The paper proposes FRAC-MARL, a decentralized actor-critic multi-agent reinforcement learning method that achieves full Byzantine resilience by leveraging redundancy in communication, ensuring convergence to optimal parameters even under adversarial attacks.