Rethinking Chain-of-Thought from the Perspective of Self-Training

TL;DR

This paper introduces a novel Chain-of-Thought framework that leverages self-training principles to enhance reasoning in large language models, addressing over-reasoning and iteration similarity issues for better performance and efficiency.

Contribution

It proposes a new CoT framework with task-specific prompting and adaptive iteration modules, improving reasoning accuracy and computational efficiency.

Findings

Significant performance improvements over existing CoT methods.

Enhanced reasoning accuracy with reduced over-reasoning.

Improved computational efficiency in reasoning processes.

Abstract

Chain-of-thought (CoT) reasoning has emerged as an effective approach for activating latent capabilities in LLMs. Interestingly, we observe that both CoT reasoning and self-training share the core objective: iteratively leveraging model-generated information to progressively reduce prediction uncertainty. Building on this insight, we propose a novel CoT framework to improve reasoning performance. Our framework integrates two key components: (i) a task-specific prompt module that optimizes the initial reasoning process, and (ii) an adaptive reasoning iteration module that dynamically refines the reasoning process and addresses the limitations of previous CoT approaches, \ie over-reasoning and high similarity between consecutive reasoning iterations. Extensive experiments demonstrate that the proposed method achieves significant advantages in both performance and computational efficiency.