A Hopfieldian View-based Interpretation for Chain-of-Thought Reasoning

TL;DR

This paper offers a Hopfieldian perspective to explain why Chain-of-Thought prompting improves reasoning in large language models, introducing a Read-and-Control method to analyze and enhance CoT accuracy across multiple datasets.

Contribution

It provides a novel Hopfieldian interpretative framework for CoT and proposes a Read-and-Control approach to understand and improve reasoning accuracy in LLMs.

Findings

Deciphers the inner workings of CoT reasoning.

Enables error localization in reasoning paths.

Improves reasoning accuracy through control mechanisms.

Abstract

Chain-of-Thought (CoT) holds a significant place in augmenting the reasoning performance for large language models (LLMs). While some studies focus on improving CoT accuracy through methods like retrieval enhancement, yet a rigorous explanation for why CoT achieves such success remains unclear. In this paper, we analyze CoT methods under two different settings by asking the following questions: (1) For zero-shot CoT, why does prompting the model with "let's think step by step" significantly impact its outputs? (2) For few-shot CoT, why does providing examples before questioning the model could substantially improve its reasoning ability? To answer these questions, we conduct a top-down explainable analysis from the Hopfieldian view and propose a Read-and-Control approach for controlling the accuracy of CoT. Through extensive experiments on seven datasets for three different tasks, we demonstrate that our framework can decipher the inner workings of CoT, provide reasoning error localization, and control to come up with the correct reasoning path.