Graph-Enhanced Large Language Models for Spatial Search

2026-06-22

AI Summary

This work addresses the limited capability of large language models (LLMs) in performing spatial reasoning tasks grounded in the physical world. To overcome this limitation, the authors propose a novel graph-augmented LLM framework that integrates spatial graph structures with retrieval-augmented generation (RAG) mechanisms, enabling joint reasoning over complex spatial data. The proposed approach significantly enhances model performance on spatial question-answering benchmarks and offers practical support for real-world applications in urban planning and civil engineering. Furthermore, it lays the groundwork for next-generation intelligent search engines capable of handling sophisticated spatial queries.

Abstract

There have been many recent improvements in the ability of Large Language Models (LLMs) to perform complex tasks and answer domain-specific questions through techniques like Retrieval Augmented Generation (RAG). However, reasoning abilities of LLMs, including spatial reasoning abilities, are still lacking. Spatial reasoning is a key component required to answer questions in a variety of domains that are grounded in the physical world, including urban planning, civil engineering, travel, and many others. To advance the development of LLMs and facilitate an impact in these domains, new research techniques must be developed to enable LLMs to reason over spatial data, which is commonly stored in the form of a graph. In this paper we outline the challenges associated with spatial reasoning through LLMs and envision a future in which search engines integrate with LLMs to answer complex spatial questions through graph-enhanced reasoning.

Problem

• spatial reasoning
• Large Language Models
• graph data
• spatial search
• physical world grounding

Innovation

• Graph-Enhanced LLMs
• Spatial Reasoning
• Retrieval Augmented Generation
• Spatial Search
• Graph-based Reasoning