The Structure-Content Trade-off in Knowledge Graph Retrieval

TL;DR

This paper investigates how different retrieval strategies affect the content and structure of knowledge graphs used by LLMs, highlighting the importance of balancing content relevance and structural coherence for better reasoning.

Contribution

It introduces a hybrid retrieval method that controls the importance of questions and subquestions, demonstrating how to optimize knowledge graph retrieval for LLM reasoning.

Findings

Subquestion-based retrieval improves content precision.

Question-based retrieval maintains structural coherence.

Optimal retrieval balances content relevance and structure.

Abstract

Large Language Models (LLMs) increasingly rely on knowledge graphs for factual reasoning, yet how retrieval design shapes their performance remains unclear. We examine how question decomposition changes the retrieved subgraph's content and structure. Using a hybrid retrieval function that controls the importance of initial question and subquestions, we show that subquestion-based retrieval improves content precision, but yields disjoint subgraphs, while question-based retrieval maintains structure at the cost of relevance. Optimal performance arises between these extremes, revealing that balancing retrieval content and structure is key to effective LLM reasoning over structured knowledge.