Skip to main navigation Skip to search Skip to main content

MITS: Enhanced Tree Search Reasoning for LLMs via Pointwise Mutual Information

Research output: Chapter in book / Conference proceedingConference article published in proceeding or bookAcademic researchpeer-review

Abstract

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 that explore multiple reasoning paths. 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.

Original languageEnglish
Title of host publicationAdvances in Knowledge Discovery and Data Mining - 30th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2026, Proceedings
EditorsRaymond Chi-Wing Wong, Hanghang Tong, Hua Lu, James Kwok, Flora Salim, Yuanfeng Song, Man Lung Yiu
PublisherSpringer Science and Business Media Deutschland GmbH
Pages288-300
Number of pages13
ISBN (Print)9789819214617
DOIs
Publication statusPublished - Jun 2026
Event30th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2026 - Hong Kong, China
Duration: 9 Jun 202612 Jun 2026

Publication series

NameLecture Notes in Computer Science
Volume16598 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference30th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2026
Country/TerritoryChina
CityHong Kong
Period9/06/2612/06/26

Keywords

  • Large Language Models
  • Reasoning

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science

Fingerprint

Dive into the research topics of 'MITS: Enhanced Tree Search Reasoning for LLMs via Pointwise Mutual Information'. Together they form a unique fingerprint.

Cite this