Unraveling the cognitive patterns of Large Language Models through module communities

TL;DR

This paper introduces a network-based framework to analyze the internal module communities of Large Language Models, revealing their emergent cognitive patterns and contrasting them with biological systems, thereby enhancing interpretability and fine-tuning strategies.

Contribution

It develops a novel network-based approach linking LLM modules, skills, and datasets, offering new insights into their cognitive organization and interpretability.

Findings

Module communities show emergent skill patterns

LLMs exhibit distributed, interconnected cognitive organization

Skill acquisition benefits from dynamic, cross-regional interactions

Abstract

Large Language Models (LLMs) have reshaped our world with significant advancements in science, engineering, and society through applications ranging from scientific discoveries and medical diagnostics to Chatbots. Despite their ubiquity and utility, the underlying mechanisms of LLM remain concealed within billions of parameters and complex structures, making their inner architecture and cognitive processes challenging to comprehend. We address this gap by adopting approaches to understanding emerging cognition in biology and developing a network-based framework that links cognitive skills, LLM architectures, and datasets, ushering in a paradigm shift in foundation model analysis. The skill distribution in the module communities demonstrates that while LLMs do not strictly parallel the focalized specialization observed in specific biological systems, they exhibit unique communities of modules whose emergent skill patterns partially mirror the distributed yet interconnected cognitive organization seen in avian and small mammalian brains. Our numerical results highlight a key divergence from biological systems to LLMs, where skill acquisition benefits substantially from dynamic, cross-regional interactions and neural plasticity. By integrating cognitive science principles with machine learning, our framework provides new insights into LLM interpretability and suggests that effective fine-tuning strategies should leverage distributed learning dynamics rather than rigid modular interventions.