Attention Heads of Large Language Models: A Survey

TL;DR

This survey systematically explores the roles and mechanisms of attention heads in large language models, introducing a four-stage framework to understand their internal reasoning processes and categorizing research methodologies and evaluation approaches.

Contribution

It presents a novel four-stage framework for analyzing attention heads and categorizes existing research methods, advancing understanding of LLM internal mechanisms.

Findings

Identification of specific functions of attention heads

Classification of research methodologies into two categories

Summary of evaluation methods and benchmarks

Abstract

Since the advent of ChatGPT, Large Language Models (LLMs) have excelled in various tasks but remain as black-box systems. Understanding the reasoning bottlenecks of LLMs has become a critical challenge, as these limitations are deeply tied to their internal architecture. Among these, attention heads have emerged as a focal point for investigating the underlying mechanics of LLMs. In this survey, we aim to demystify the internal reasoning processes of LLMs by systematically exploring the roles and mechanisms of attention heads. We first introduce a novel four-stage framework inspired by the human thought process: Knowledge Recalling, In-Context Identification, Latent Reasoning, and Expression Preparation. Using this framework, we comprehensively review existing research to identify and categorize the functions of specific attention heads. Additionally, we analyze the experimental methodologies used to discover these special heads, dividing them into two categories: Modeling-Free and Modeling-Required methods. We further summarize relevant evaluation methods and benchmarks. Finally, we discuss the limitations of current research and propose several potential future directions.