[论文解读] DA-RAG: Dynamic Attributed Community Search for Retrieval-Augmented Generation

Xingyuan Zeng, Zuohan Wu+5 more|arXiv (Cornell University)|Feb 9, 2026

Information Retrieval and Search Behavior被引用 0

一句话总结

DA-RAG 引入 Embedding-Attributed Community Search (EACS) 和 chunk-layer 图索引，以动态检索高阶、面向查询的子图用于 RAG，在准确性和效率方面优于基线。

ABSTRACT

Owing to their unprecedented comprehension capabilities, large language models (LLMs) have become indispensable components of modern web search engines. From a technical perspective, this integration represents retrieval-augmented generation (RAG), which enhances LLMs by grounding them in external knowledge bases. A prevalent technical approach in this context is graph-based RAG (G-RAG). However, current G-RAG methodologies frequently underutilize graph topology, predominantly focusing on low-order structures or pre-computed static communities. This limitation affects their effectiveness in addressing dynamic and complex queries. Thus, we propose DA-RAG, which leverages attributed community search (ACS) to extract relevant subgraphs based on the queried question dynamically. DA-RAG captures high-order graph structures, allowing for the retrieval of self-complementary knowledge. Furthermore, DA-RAG is equipped with a chunk-layer oriented graph index, which facilitates efficient multi-granularity retrieval while significantly reducing both computational and economic costs. We evaluate DA-RAG on multiple datasets, demonstrating that it outperforms existing RAG methods by up to 40% in head-to-head comparisons across four metrics while reducing index construction time and token overhead by up to 37% and 41%, respectively.

研究动机与目标

• 通过利用静态社区之外的高阶图结构来提升 RAG 的检索效果。
• 提出 Embedding-Attributed Community Search，在动态提取结构上紧密、与查询相关的子图。
• 设计一个面向块层的图索引，以实现高效的多粒度检索，而不依赖于大量聚类。
• 实证显示 DA-RAG 的性能优于基线并降低索引与标记成本。

提出的方法

• 定义 Embedding-Attributed Community Search (EACS)，在 k-truss 连通子图中最大化查询相关性，缓解自由乘客效应。
• 开发三层离线索引：语义块层 (L_C)、知识图谱层 (L_KG) 与相似度层 (L_S)；通过层间和层内边连接各层。
• 使用自粗到细的在线检索：先在 L_C 识别 H_C，再剪枝为 G_KG^work 与 G_S^work，最后在精 refined 子图上应用 EACS 获得 H_KG 与 H_S。
• 通过候选评估循环结合大模型辅助评分与上下文预算约束，动态确定 k-truss 的 k 值。
• 提出 Q-Peel 启发式方法以高效解决 EACS，给出 NP-hard 性质与分析复杂度。
• 提供在 UltraDomain 数据集上对多基线的端到端评估，衡量有效性和效率。

Figure 1. Differences between existing methods and our method. (a) Methods w/o community concern are limited to low-order graph topology, capturing only partial aspects. (b) Methods with static community partition could return a diverging and unfocused response. (c) Our method retrieves a query-rele

实验结果

研究问题

• RQ1 与最先进基线相比，DA-RAG 在检索质量和答案准确性方面的表现如何？
• RQ2 DA-RAG 在索引构建和在线检索方面有多高的效率？
• RQ3 DA-RAG 检索得到的子图在结构凝聚性和语义相关性方面是否优于其他方法？
• RQ4 EACS 的自适应 k 确定对不同查询的性能有何影响？

主要发现

• DA-RAG 在四个指标的头对头比较中领先基线，优势高达 40%。
• 索引构建时间和标记开销分别降低至原来的最高 37% 和 41%。
• 在线检索阶段，标记消耗平均下降 73.8%（在某些数据集上最高达 88.76%），延迟与 GraphRAG-Global 相当。
• DA-RAG 采用自粗到细的检索策略和 EACS，产出高质量、具凝聚性的子图。
• 自适应的 k 确定过程与 Q-Peel 启发式方法提供高效、面向查询的子图提取，具有可证明的凝聚性和受限的推理跳数。

Figure 2. Overview of the DA-RAG framework: (a) Offline Indexing creates a novel graph index from source documents, comprising a high-level layer ( $L_{C}$ ) and two granular layers ( $L_{KG}$ and $L_{S}$ ). (b) Online Retrieval employs a coarse-to-fine strategy. (c) EACS Formulation defines the sub