COA-Bench provides a benchmark for comparing course-of-action generation policies using self-play and game-theoretic algorithms. It evaluates typed action chains through BLUE-vs-RED advantage scores and Nash-gap distance to measure tactical coherence against adversarial responses.
HOW THIS AFFECTS YOU
●
researcherYou can use this artifact to evaluate multi-agent systems in distributed planning and adversarial environments.