Self-Evaluation Guided Beam Search for Reasoning

–Neural Information Processing Systems 

Breaking down a problem into intermediate steps has demonstrated impressive performance in Large Language Model (LLM) reasoning.