Saved in:
Bibliographic Details
Main Authors: Li, Jiaxi, Shi, Yucheng, Huang, Xiao, Lu, Jin, Liu, Ninghao
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.03632
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910034523324416
author Li, Jiaxi
Shi, Yucheng
Huang, Xiao
Lu, Jin
Liu, Ninghao
author_facet Li, Jiaxi
Shi, Yucheng
Huang, Xiao
Lu, Jin
Liu, Ninghao
contents Tree search has become as a representative framework for test-time reasoning with large language models (LLMs), exemplified by methods such as Tree-of-Thought and Monte Carlo Tree Search. However, it remains difficult to provide instant and reliable quantitative assessments of intermediate reasoning step quality, and extensive path exploration is computationally costly. To address this, we propose Mutual Information Tree Search (MITS), a novel framework that guides reasoning with information-theoretic principles. MITS introduces an effective scoring function based on pointwise mutual information (PMI), which enables step-wise evaluation of reasoning paths and search tree expansion via beam search without expensive look-ahead simulations, achieving superior reasoning performances while maintaining computational efficiency. The framework is complemented by an entropy-based dynamic sampling strategy that adaptively allocates computational resources to uncertain reasoning steps where exploration is most beneficial. For final prediction, MITS employs a weighted voting scheme that combines PMI scores with prediction consensus. Through comprehensive experiments on diverse reasoning benchmarks, MITS consistently surpasses baseline methods, establishing a principled and efficient framework for LLM reasoning. The code is available at https://github.com/plusnli/MITS.
format Preprint
id arxiv_https___arxiv_org_abs_2510_03632
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MITS: Enhanced Tree Search Reasoning for LLMs via Pointwise Mutual Information
Li, Jiaxi
Shi, Yucheng
Huang, Xiao
Lu, Jin
Liu, Ninghao
Artificial Intelligence
Tree search has become as a representative framework for test-time reasoning with large language models (LLMs), exemplified by methods such as Tree-of-Thought and Monte Carlo Tree Search. However, it remains difficult to provide instant and reliable quantitative assessments of intermediate reasoning step quality, and extensive path exploration is computationally costly. To address this, we propose Mutual Information Tree Search (MITS), a novel framework that guides reasoning with information-theoretic principles. MITS introduces an effective scoring function based on pointwise mutual information (PMI), which enables step-wise evaluation of reasoning paths and search tree expansion via beam search without expensive look-ahead simulations, achieving superior reasoning performances while maintaining computational efficiency. The framework is complemented by an entropy-based dynamic sampling strategy that adaptively allocates computational resources to uncertain reasoning steps where exploration is most beneficial. For final prediction, MITS employs a weighted voting scheme that combines PMI scores with prediction consensus. Through comprehensive experiments on diverse reasoning benchmarks, MITS consistently surpasses baseline methods, establishing a principled and efficient framework for LLM reasoning. The code is available at https://github.com/plusnli/MITS.
title MITS: Enhanced Tree Search Reasoning for LLMs via Pointwise Mutual Information
topic Artificial Intelligence
url https://arxiv.org/abs/2510.03632