Millions of $\text{GeAR}$-s: Extending GraphRAG to Millions of Documents

TL;DR

This paper adapts the GeAR graph-based retrieval-augmented generation method to handle millions of documents, testing its scalability and effectiveness across broader datasets beyond specific tasks.

Contribution

It extends the GeAR approach to large-scale datasets, demonstrating its potential for general-purpose retrieval-augmented generation.

Findings

Scalable to millions of documents

Effective across diverse datasets

Identifies limitations of current graph-based RAG methods

Abstract

Recent studies have explored graph-based approaches to retrieval-augmented generation, leveraging structured or semi-structured information -- such as entities and their relations extracted from documents -- to enhance retrieval. However, these methods are typically designed to address specific tasks, such as multi-hop question answering and query-focused summarisation, and therefore, there is limited evidence of their general applicability across broader datasets. In this paper, we aim to adapt a state-of-the-art graph-based RAG solution: $\text{GeAR}$ and explore its performance and limitations on the SIGIR 2025 LiveRAG Challenge.