Token Signature: Predicting Chain-of-Thought Gains with Token Decoding Feature in Large Language Models

TL;DR

This paper introduces a novel method called Dynamic CoT that uses token probability distribution indicators to assess and improve chain-of-thought reasoning in large language models, reducing token usage while maintaining accuracy.

Contribution

It proposes token distribution-based indicators for evaluating CoT effectiveness and develops Dynamic CoT, a dynamic selection method that enhances reasoning efficiency in LLMs.

Findings

Indicators achieve 89.2% accuracy in assessing CoT effectiveness.

Dynamic CoT reduces token consumption by over 35%.

Method extends to closed-source models via transfer learning.

Abstract

Chain-of-Thought (CoT) technique has proven effective in improving the performance of large language models (LLMs) on complex reasoning tasks. However, the performance gains are inconsistent across different tasks, and the underlying mechanism remains a long-standing research question. In this work, we make a preliminary observation that the monotonicity of token probability distributions may be correlated with the gains achieved through CoT reasoning. Leveraging this insight, we propose two indicators based on the token probability distribution to assess CoT effectiveness across different tasks. By combining instance-level indicators with logistic regression model, we introduce Dynamic CoT, a method that dynamically select between CoT and direct answer. Furthermore, we extend Dynamic CoT to closed-source models by transferring decision strategies learned from open-source models. Our indicators for assessing CoT effectiveness achieve an accuracy of 89.2\%, and Dynamic CoT reduces token consumption by more than 35\% while maintaining high accuracy. Overall, our work offers a novel perspective on the underlying mechanisms of CoT reasoning and provides a framework for its more efficient deployment.