Agentic graphrag uses symbolic knowledge to improve language model answers

Typed-Centrality Grounding for Agentic GraphRAG: A Symbolic Approach for Agentic Reasoning

Artificial Intelligence

Summary

Large language models can give better answers if they use clear, exact knowledge from structured sources like knowledge graphs. The authors built a system that gives language models this exact knowledge by looking at the importance of different types of things in a knowledge graph separately. This helps the system reason more accurately instead of just guessing from text similarities. Tests showed the system gave answers that fit the facts better with this approach.

What this means in practice

  • For enterprise search teams: Improve context consistency in knowledge-based search by integrating typed centrality symbolic reasoning with language models.
  • For intelligent assistant developers: Enhance agentic reasoning in assistants by using symbolic knowledge from typed centralities in knowledge graphs to support language model outputs.

Tested on one dataset.

Authors

Rosario Napoli, Antonio Celesti, Massimo Villari, Maria Fazio

Abstract

Knowledge Graphs (KGs) and Large Language Models (LLMs) represent two approaches for knowledge representation and reasoning over heterogeneous data. Recent research has explored their integration through Graph Retrieval-Augmented Generation (GraphRAG) systems, which use KGs to provide structured context to LLMs. However, existing GraphRAG approaches treat KGs as homogeneous graphs and rely on textual similarity between the input prompt and the context, rather than using it as a source of deterministic knowledge. This work proposes an agent-based GraphRAG system that introduces a symbolic layer to integrate deterministic reasoning alongside text similarities. This approach is based on the definition of typed centralities for KGs, where centrality measures are computed separately for each specific KG's class. This information is used to construct a symbolic representation that is provided to the LLM as a deterministic source. The proposed system was mathematically validated and evaluated on a real-world POLE (Person, Object, Location, Event) dataset, comparing reasoning and information retrieval with and without the symbolic layer. The results indicate that the our approach leads to responses that are more consistent with the context, improving the generated outputs.