Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FG-RAG: Enhancing Query-Focused Summarization with Context-Aware Fine-Grained Graph RAG

Yubin Hong, Chaofan Li, Jingyi Zhang, Yingxia Shao

TL;DR

FG-RAG tackles two key deficiencies of GraphRAG in query-focused summarization: insufficient contextual coverage during graph retrieval and a lack of query awareness in summarization. It proposes Context-Aware Entity Expansion to broaden contextual retrieval and Query-Level Fine-Grained Summarization to produce per-entity, query-aligned summaries that are consolidated into a final answer. The approach achieves higher comprehensiveness, diversity, and empowerment on QFS benchmarks and maintains favorable token efficiency, across multiple LLMs and domains. This work advances retrieval-augmented generation for domain-specific, multi-hop reasoning tasks and offers practical guidance for building context-rich, query-aware graphs for LLMs.

Abstract

Retrieval-Augmented Generation (RAG) enables large language models to provide more precise and pertinent responses by incorporating external knowledge. In the Query-Focused Summarization (QFS) task, GraphRAG-based approaches have notably enhanced the comprehensiveness and diversity of generated responses. However, existing GraphRAG-based approaches predominantly focus on coarse-grained information summarization without being aware of the specific query, and the retrieved content lacks sufficient contextual information to generate comprehensive responses. To address the deficiencies of current RAG systems, we propose Context-Aware Fine-Grained Graph RAG (FG-RAG) to enhance the performance of the QFS task. FG-RAG employs Context-Aware Entity Expansion in graph retrieval to expand the coverage of retrieved entities in the graph, thus providing enough contextual information for the retrieved content. Furthermore, FG-RAG utilizes Query-Level Fine-Grained Summarization to incorporate fine-grained details during response generation, enhancing query awareness for the generated summarization. Our evaluation demonstrates that FG-RAG outperforms other RAG systems in multiple metrics of comprehensiveness, diversity, and empowerment when handling the QFS task. Our implementation is available at https://github.com/BuptWululu/FG-RAG.

FG-RAG: Enhancing Query-Focused Summarization with Context-Aware Fine-Grained Graph RAG

TL;DR

FG-RAG tackles two key deficiencies of GraphRAG in query-focused summarization: insufficient contextual coverage during graph retrieval and a lack of query awareness in summarization. It proposes Context-Aware Entity Expansion to broaden contextual retrieval and Query-Level Fine-Grained Summarization to produce per-entity, query-aligned summaries that are consolidated into a final answer. The approach achieves higher comprehensiveness, diversity, and empowerment on QFS benchmarks and maintains favorable token efficiency, across multiple LLMs and domains. This work advances retrieval-augmented generation for domain-specific, multi-hop reasoning tasks and offers practical guidance for building context-rich, query-aware graphs for LLMs.

Abstract

Retrieval-Augmented Generation (RAG) enables large language models to provide more precise and pertinent responses by incorporating external knowledge. In the Query-Focused Summarization (QFS) task, GraphRAG-based approaches have notably enhanced the comprehensiveness and diversity of generated responses. However, existing GraphRAG-based approaches predominantly focus on coarse-grained information summarization without being aware of the specific query, and the retrieved content lacks sufficient contextual information to generate comprehensive responses. To address the deficiencies of current RAG systems, we propose Context-Aware Fine-Grained Graph RAG (FG-RAG) to enhance the performance of the QFS task. FG-RAG employs Context-Aware Entity Expansion in graph retrieval to expand the coverage of retrieved entities in the graph, thus providing enough contextual information for the retrieved content. Furthermore, FG-RAG utilizes Query-Level Fine-Grained Summarization to incorporate fine-grained details during response generation, enhancing query awareness for the generated summarization. Our evaluation demonstrates that FG-RAG outperforms other RAG systems in multiple metrics of comprehensiveness, diversity, and empowerment when handling the QFS task. Our implementation is available at https://github.com/BuptWululu/FG-RAG.

Paper Structure

This paper contains 25 sections, 4 equations, 4 figures, 10 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Chunk-based Retrieval-Augmented Generation
  4. Graph-based Retrieval-Augmented Generation
  5. FG-RAG Framework
  6. Graph Index
  7. Graph Retrieval with Context-Enhanced Entity Expansion
  8. Query-Level Fine-Grained Summarization
  9. Experiments
  10. Experimental Settings
  11. Overall Performance on the QFS Task
  12. Overall Performance on the Multi-hop QA Task
  13. Ablation Studies
  14. Case Study
  15. Comparison of Method Efficiency
  16. ...and 10 more sections

Figures (4)

  • Figure 1: The overall workflow of FG-RAG. In Query-Level Fine-Grained Summarization, for each entity mentioned in the query, we sequentially perform the tasks of asking relevant questions, retrieving pertinent descriptions, and generating summaries. These summaries are then provided to the LLM, which uses them to formulate the answer.
  • Figure 2: Prompts for query entities extraction.
  • Figure 3: Prompts for relevant questions formulation.
  • Figure 4: Prompts for summaries generation.