问题树：通过组合性改进结构化问题解决

Tree of Problems: Improving structured problem solving with compositionality

October 9, 2024

作者: Armel Zebaze, Benoît Sagot, Rachel Bawden

cs.AI

摘要

大型语言模型（LLMs）通过上下文学习展现出在多个任务上的显著性能。对于需要逐步思考的复杂推理任务，思维链（CoT）提示在与自一致性相结合时表现出令人印象深刻的结果。然而，一些任务对LLMs来说仍然特别难以解决。思维树（ToT）和思维图（GoT）作为替代方案出现，将复杂问题分解为子问题路径。在本文中，我们提出了问题树（ToP），这是ToT的简化版本，我们假设它可以更好地处理可以分解为相同子任务的复杂任务。我们的实证结果表明，我们的方法在复杂推理任务上的表现优于ToT和GoT，并且比CoT表现更好。本文所有代码均可在以下网址公开获取：https://github.com/ArmelRandy/tree-of-problems。

English

Large Language Models (LLMs) have demonstrated remarkable performance across multiple tasks through in-context learning. For complex reasoning tasks that require step-by-step thinking, Chain-of-Thought (CoT) prompting has given impressive results, especially when combined with self-consistency. Nonetheless, some tasks remain particularly difficult for LLMs to solve. Tree of Thoughts (ToT) and Graph of Thoughts (GoT) emerged as alternatives, dividing the complex problem into paths of subproblems. In this paper, we propose Tree of Problems (ToP), a simpler version of ToT, which we hypothesise can work better for complex tasks that can be divided into identical subtasks. Our empirical results show that our approach outperforms ToT and GoT, and in addition performs better than CoT on complex reasoning tasks. All code for this paper is publicly available here: https://github.com/ArmelRandy/tree-of-problems.