BARD-MARL

New framework verifies safety of learned multi-agent communication policies

Researchers have developed a novel framework for formally verifying the safety of learned communication policies in multi-agent reinforcement learning (MARL) systems. This approach distills complex neural policies into …