Self-Consistency：Self-Consistency Improves Chain of Thought Reasoning in Language Models

arXiv · Google Research, Brain Team

提出基于采样和多数投票的自洽解码策略，显著提升大语言模型复杂推理的准确率。