Zoom introduces a reinforcement learning framework that replaces static preference triplets with continuous, high-quality ...