Taming the Long-Tail: Efficient Reasoning RL Training with Adaptive Drafter - Explained Simply | ArXiv Explained