Landscape of Thoughts: Visualizing the Reasoning Process of Large Language Models

2025-03-28

AI Summary

Existing research lacks a comprehensive understanding of chain-of-thought (CoT) reasoning in large language models (LLMs), hindering interpretability and safety evaluation. To address this, we propose the *Thought Landscape*—a novel conceptual framework—and introduce the first interpretable, visualization-oriented framework for multiple-choice tasks. It maps CoT reasoning paths onto answer-space feature vectors and projects them into a 2D geometric landscape via t-SNE. The framework enables unified geometric analysis across models, tasks, and answer options, and integrates a lightweight validator interface. Experiments demonstrate its effectiveness in distinguishing strong versus weak models, correct versus incorrect answers, and distinct reasoning tasks, while revealing anomalous patterns such as low consistency and high uncertainty. The implementation is open-sourced and has garnered significant community attention.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Computer Vision: Large Vision Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Numerous applications of large language models (LLMs) rely on their ability to perform step-by-step reasoning. However, the reasoning behavior of LLMs remains poorly understood, posing challenges to research, development, and safety. To address this gap, we introduce landscape of thoughts-the first visualization tool for users to inspect the reasoning paths of chain-of-thought and its derivatives on any multi-choice dataset. Specifically, we represent the states in a reasoning path as feature vectors that quantify their distances to all answer choices. These features are then visualized in two-dimensional plots using t-SNE. Qualitative and quantitative analysis with the landscape of thoughts effectively distinguishes between strong and weak models, correct and incorrect answers, as well as different reasoning tasks. It also uncovers undesirable reasoning patterns, such as low consistency and high uncertainty. Additionally, users can adapt our tool to a model that predicts the property they observe. We showcase this advantage by adapting our tool to a lightweight verifier that evaluates the correctness of reasoning paths. The code is publicly available at: https://github.com/tmlr-group/landscape-of-thoughts.

Problem

• Visualizing LLM reasoning paths for better understanding
• Distinguishing model performance and answer correctness
• Identifying undesirable reasoning patterns in LLMs

Innovation

• Visualizes reasoning paths using feature vectors
• Employs t-SNE for 2D plotting of reasoning states
• Adaptable tool for model property prediction