蒙特卡罗规划的忠实问答（英文版）

英文标题：Faithful Question Answering with Monte-Carlo Planning中文摘要：本论文提出 FAithful question answering with MontE-carlo planning （FAME）方法，以结构化蕴含树方式展示中间推理步骤，包含多个基础任务模块并通过蒙特卡罗规划算法作为控制器提出行动，旨在回答问题并与大型语言模型相比，使用更小的模型大小实现了最先进的性能。英文摘要：Although large language models demonstrate remarkable question-answeringperformances, revealing the intermediate reasoning steps that the modelsfaithfully follow remains challenging. In this paper, we propose FAME (FAithfulquestion answering with MontE-carlo planning) to answer