A Graph Perspective to Probe Structural Patterns of Knowledge in Large Language Models
This work addresses the gap in analyzing structural knowledge patterns in LLMs, which is incremental as it builds on existing knowledge base studies but introduces a novel graph-based approach.
The paper tackles the problem of understanding structural patterns of knowledge in large language models by analyzing them from a graph perspective, uncovering knowledge homophily and developing a graph model to estimate entity knowledge, which when used for fine-tuning leads to superior performance.
Large language models have been extensively studied as neural knowledge bases for their knowledge access, editability, reasoning, and explainability. However, few works focus on the structural patterns of their knowledge. Motivated by this gap, we investigate these structural patterns from a graph perspective. We quantify the knowledge of LLMs at both the triplet and entity levels, and analyze how it relates to graph structural properties such as node degree. Furthermore, we uncover the knowledge homophily, where topologically close entities exhibit similar levels of knowledgeability, which further motivates us to develop graph machine learning models to estimate entity knowledge based on its local neighbors. This model further enables valuable knowledge checking by selecting triplets less known to LLMs. Empirical results show that using selected triplets for fine-tuning leads to superior performance.