UserRL: Training Interactive User-Centric Agent via Reinforcement Learning Paper • 2509.19736 • Published Sep 24 • 11