Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CG-RAG: Research Question Answering by Citation Graph Retrieval-Augmented LLMs

Yuntong Hu, Zhihan Lei, Zhongjie Dai, Allen Zhang, Abhinav Angirekula, Zheng Zhang, Liang Zhao

TL;DR

The paper tackles research question answering over scientific literature by leveraging citation graphs to capture inter-paper relationships. It introduces Lexical-Semantic Graph Retrieval (LeSeGR), a graph-encoder-based fusion of sparse lexical and dense semantic signals, and Contextualized Graph Retrieval-Augmented Generation (CG-RAG) that uses top-N contextual subgraphs to guide generation. By formalizing a contextual, chunk-level citation graph and propagating information through an entangled retrieval framework, the authors demonstrate state-of-the-art retrieval and generation performance on PubMedQA-1k and PapersWithCodeQA, along with meaningful efficiency gains. The work highlights the practical impact of graph-contextualized representations for scalable, accurate scholarly question answering, and generalizes beyond existing retrieval paradigms to integrate structure and semantics in literature graphs.

Abstract

Research question answering requires accurate retrieval and contextual understanding of scientific literature. However, current Retrieval-Augmented Generation (RAG) methods often struggle to balance complex document relationships with precise information retrieval. In this paper, we introduce Contextualized Graph Retrieval-Augmented Generation (CG-RAG), a novel framework that integrates sparse and dense retrieval signals within graph structures to enhance retrieval efficiency and subsequently improve generation quality for research question answering. First, we propose a contextual graph representation for citation graphs, effectively capturing both explicit and implicit connections within and across documents. Next, we introduce Lexical-Semantic Graph Retrieval (LeSeGR), which seamlessly integrates sparse and dense retrieval signals with graph encoding. It bridges the gap between lexical precision and semantic understanding in citation graph retrieval, demonstrating generalizability to existing graph retrieval and hybrid retrieval methods. Finally, we present a context-aware generation strategy that utilizes the retrieved graph-structured information to generate precise and contextually enriched responses using large language models (LLMs). Extensive experiments on research question answering benchmarks across multiple domains demonstrate that our CG-RAG framework significantly outperforms RAG methods combined with various state-of-the-art retrieval approaches, delivering superior retrieval accuracy and generation quality.

CG-RAG: Research Question Answering by Citation Graph Retrieval-Augmented LLMs

TL;DR

The paper tackles research question answering over scientific literature by leveraging citation graphs to capture inter-paper relationships. It introduces Lexical-Semantic Graph Retrieval (LeSeGR), a graph-encoder-based fusion of sparse lexical and dense semantic signals, and Contextualized Graph Retrieval-Augmented Generation (CG-RAG) that uses top-N contextual subgraphs to guide generation. By formalizing a contextual, chunk-level citation graph and propagating information through an entangled retrieval framework, the authors demonstrate state-of-the-art retrieval and generation performance on PubMedQA-1k and PapersWithCodeQA, along with meaningful efficiency gains. The work highlights the practical impact of graph-contextualized representations for scalable, accurate scholarly question answering, and generalizes beyond existing retrieval paradigms to integrate structure and semantics in literature graphs.

Abstract

Research question answering requires accurate retrieval and contextual understanding of scientific literature. However, current Retrieval-Augmented Generation (RAG) methods often struggle to balance complex document relationships with precise information retrieval. In this paper, we introduce Contextualized Graph Retrieval-Augmented Generation (CG-RAG), a novel framework that integrates sparse and dense retrieval signals within graph structures to enhance retrieval efficiency and subsequently improve generation quality for research question answering. First, we propose a contextual graph representation for citation graphs, effectively capturing both explicit and implicit connections within and across documents. Next, we introduce Lexical-Semantic Graph Retrieval (LeSeGR), which seamlessly integrates sparse and dense retrieval signals with graph encoding. It bridges the gap between lexical precision and semantic understanding in citation graph retrieval, demonstrating generalizability to existing graph retrieval and hybrid retrieval methods. Finally, we present a context-aware generation strategy that utilizes the retrieved graph-structured information to generate precise and contextually enriched responses using large language models (LLMs). Extensive experiments on research question answering benchmarks across multiple domains demonstrate that our CG-RAG framework significantly outperforms RAG methods combined with various state-of-the-art retrieval approaches, delivering superior retrieval accuracy and generation quality.
Paper Structure (18 sections, 1 theorem, 11 equations, 2 figures, 5 tables, 1 algorithm)

This paper contains 18 sections, 1 theorem, 11 equations, 2 figures, 5 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Information Retrieval
  4. Retrieval-Augmented Generation
  5. Research Question Answering
  6. Methodology
  7. Overview
  8. Contextual Citation Graph
  9. Chunk Node
  10. Chunk-Chunk Edge
  11. Lexical-Semantic Graph Retrieval (LeSeGR)
  12. Contextualized Graph Retrieval-Augmented Generation (CG-RAG)
  13. Experiment
  14. Settings
  15. Main Results
  16. ...and 3 more sections

Key Result

proposition 1

Post-retrieval methods with the metric $f_{\text{hybrid}}: f_{\text{sparse}} \bigoplus f_{\text{dense}} \to \mathbb{R}$, represent a special case of the proposed Lexical-Semantic Graph Retrieval (LeSeGR). This holds when no additional relevant contextual information exists for any chunk in the citat

Figures (2)

  • Figure 1: Illustration of retrieval-augmented research question answering using: (a) sparse retrieval based on lexical matches, (b) dense retrieval based on semantic relevance, and (c) contextualized retrieval leveraging graph context, i.e., interactions between documents. represents the dense embedding, where deeper colors indicate a higher semantic relevance to the question. Boxed text in red highlights the matched terms between questions and documents, while Boxed text in blue highlights the matched terms between documents.
  • Figure 2: Overview of Contextualized Graph Retrieval-Augmented Generation.

Theorems & Definitions (1)

  • proposition 1: LeSeGR Generality