Accelerating RL for LLM Reasoning with Optimal Advantage Regression | Read Paper on Bytez