Understanding Reasoning in Chain-of-Thought from the Hopfieldian View

TL;DR

This paper introduces a Hopfieldian perspective to understand Chain-of-Thought reasoning in large language models, connecting cognitive elements to improve robustness and interpretability of reasoning processes.

Contribution

It presents a novel cognitive framework for CoT reasoning, a method for localizing reasoning errors, and the Representation-of-Thought (RoT) framework to enhance robustness.

Findings

RoT improves reasoning robustness

RoT enhances interpretability of CoT

Method localizes reasoning errors effectively

Abstract

Large Language Models have demonstrated remarkable abilities across various tasks, with Chain-of-Thought (CoT) prompting emerging as a key technique to enhance reasoning capabilities. However, existing research primarily focuses on improving performance, lacking a comprehensive framework to explain and understand the fundamental factors behind CoT's success. To bridge this gap, we introduce a novel perspective grounded in the Hopfieldian view of cognition in cognitive neuroscience. We establish a connection between CoT reasoning and key cognitive elements such as stimuli, actions, neural populations, and representation spaces. From our view, we can understand the reasoning process as the movement between these representation spaces. Building on this insight, we develop a method for localizing reasoning errors in the response of CoTs. Moreover, we propose the Representation-of-Thought (RoT) framework, which leverages the robustness of low-dimensional representation spaces to enhance the robustness of the reasoning process in CoTs. Experimental results demonstrate that RoT improves the robustness and interpretability of CoT reasoning while offering fine-grained control over the reasoning process.