Stable Reinforcement Learning for Efficient Reasoning | Read Paper on Bytez