Parrot Mind: Towards Explaining the Complex Task Reasoning of Pretrained Large Language Models with Template-Content Structure

TL;DR

This paper introduces the template-content (T-C) structure as an intrinsic constraint in large language models that explains their ability to perform complex multi-step reasoning efficiently, supported by theory and experiments.

Contribution

The paper proposes the T-C structure as a key to understanding and improving reasoning in LLMs, extending it to hierarchical forms for multi-step tasks, with theoretical and experimental validation.

Findings

T-C structure reduces reasoning space from exponential to linear.

Hierarchical T-C structure enables models to learn multi-step reasoning efficiently.

Experimental validation shows T-C structure exists in current LLMs and aids reasoning.

Abstract

The pre-trained large language models (LLMs) have shown their extraordinary capacity to solve reasoning tasks, even on tasks that require a complex process involving multiple sub-steps. However, given the vast possible generation space of all the tasks, how the pretrained model learns the reasoning ability remains an open question. We firstly propose that an intrinsic structural constraint on the generated sequence of language-based reasoning -- we called it template-content structure (T-C structure) -- is the key to explain why LLMs can solve a large number of complex reasoning problems with limited training data by showing this structure can reduce the possible space from exponential level to linear level. Furthermore, by generalizing this structure to the hierarchical case, we demonstrate that models can achieve task composition, further reducing the space needed to learn from linear to logarithmic, thereby effectively learning on complex reasoning involving multiple steps. We provide both examples and formal theory of our T-C structure. We also experimentally validate the existence of the T-C structure in some current LLMs and its effectiveness for reasoning.