The Graph Language: How Knowledge Graphs Speak to Large Language Models

2026-08-02

AI Summary

This work addresses the semantic gap in integrating knowledge graphs (KGs) with large language models (LLMs) by proposing the GRALAN framework. GRALAN introduces a trainable linguistic intermediary module that translates KG relational structures into relation-aware semantic tokens, enabling structured knowledge to be directly “spoken” within the LLM’s semantic space—without fine-tuning the LLM parameters. By reframing question answering as a subgraph-oriented entity classification task, GRALAN preserves graph structural fidelity while substantially enhancing reasoning capabilities. Experimental results demonstrate that the framework significantly outperforms existing methods on multi-hop complex question answering benchmarks, confirming its effectiveness in knowledge-intensive scenarios and establishing a new paradigm for KG–LLM integration.

Abstract

Large Language Models (LLMs) excel at reasoning but benefit from grounding provided by Knowledge Graphs (KGs). However, integrating these paradigms is challenging. We introduce GRALAN, which enables KGs to speak directly in the LLM's semantic space through relational tokens that preserve graph structure. GRALAN-s trainable language mediator generates structured tokens for any frozen LLM, creating a foundation for knowledge-intensive applications. We demonstrate its effectiveness in question-answering by re-framing the task as entity classification over question-focused subgraphs. Experiments show that GRALAN significantly outperforms existing methods, particularly on complex multi-hop reasoning tasks, establishing a new paradigm for KG-LLM integration that maintains structural fidelity while leveraging LLMs' reasoning capabilities.

Problem

• Knowledge Graphs
• Large Language Models
• Graph-LLM Integration
• Structural Fidelity
• Multi-hop Reasoning

Innovation

• Knowledge Graphs
• Large Language Models
• Relational Tokens
• Structured Representation
• Multi-hop Reasoning