MetaAgent-X : Breaking the Ceiling of Automatic Multi-Agent Systems via End-to-End Reinforcement Learning Paper • 2605.14212 • Published May 14 • 19