Exploring the Taxonomy of Survey Papers on Large Language Models Using Classical Machine Learning

Abstract

The rapid advancements in large language models (LLMs) have resulted in an exponential growth of survey papers. To manage and understand the evolving taxonomy of these surveys, this project employs graph representation learning including with classical Machine learning algorithms. By treating survey topics and their interrelationships as a graph, we explore the evolving taxonomy in the context of large language models. Our study highlights trends in survey papers, providing a comprehensive understanding of their distribution across various research domains. The results reveal a clear trajectory of increasing specialization, emphasizing the role of LLMs in areas such as prompt engineering, multimodal models, and application domains like education and finance. Graph-based analysis allows us to capture these trends more effectively and enables a better understanding of the survey landscape.