PEA: Enhancing LLM Performance on Computational-Reasoning Tasks