[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86159-en":3,"doc-seo-86159-105":30,"detail-sidebar-cat-0-en-105":92},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},86159,962075114101,"Seraphina","https://ap-avatar.wpscdn.com/avatar/e000253a75eb197efd?x-image-process=image/resize,m_fixed,w_180,h_180&k=1780044092746381165",8,"Research & Report","NGM RAG Neural Graph Matching Based Retrieval Augmented Generation","Retrieval-Augmented Generation (RAG) improves Large Language Models by integrating external databases, but traditional text-based retrieval often fails on multi-hop questions where correct answers require relational reasoning. NGM-RAG introduces a unified framework that explicitly performs graph construction, neural graph matching, and answer generation. It combines text matching with Graph Neural Networks and uses adaptive weighting to select the most relevant contextual nodes. Experiments on multi-hop question answering and long-context summarization show stronger results than NaiveRAG and graph-enhanced baselines like GraphRAG and LightRAG.","NGM-RAG: Neural Graph Matching based Retrieval-Augmented  \nGeneration  \nGuo Chen1,†, Ziwen Li1,†, Maolin Zheng1 , Hao Gao2 , Junjie Huang1,* , Tao Jia1  \n1 College of Computer and Information Science, Southwest University, China  \n2Beijing Institute of Control Engineering, China  \n{cg1281838223, pique0202, [zml778922461}@email.swu.edu.cn](zml778922461}@email.swu.edu.cn)  \n[goleey@163.com](goleey@163.com) , [junjiehuang@swu.edu.cn](junjiehuang@swu.edu.cn) , [tjia@swu.edu.cn](tjia@swu.edu.cn)  \n†Equal contribution. * Corresponding author.  \narXiv :2607 . 1 1 159v 1 [ cs .IR] 13 Jul 2026  \nAbstract  \nRetrieval-Augmented Generation (RAG) significantly enhances the ability of Large Language Models (LLMs) to provide accurate and contextually relevant answers by dynamically integrating external databases. However, traditional RAG methods are primarily constrained by their reliance on text-based retrieval strategies, which often struggle with complex questions requiring multi-hop reasoning. To address this limitation, we introduce Neural Graph Matching based Retrieval-Augmented Generation (NGM-RAG), a novel framework that leverages graph structures to effectively capture and utilize relational knowledge for improved retrieval and answer generation. NGMRAG explicitly incorporates graph construction, graph matching, and answer generation into a unified process. Within this framework, we propose a neural graph matching approach that combines text-based matching with Graph Neural Networks (GNNs) . By employing an adaptive weighting strategy, NGM-RAG efficiently integrates multiple matching methods to select the most relevant contextual node information for answer generation. Experimental results on multi-hop question answering and longcontext summarization tasks demonstrate that our NGM-RAG model achieves superior performance compared to both traditional NaiveRAG methods and state-of-the-art graph-enhanced approaches such as GraphRAG and LightRAG.  \n1 Introduction  \nRetrieval-Augmented Generation (RAG) is an innovative approach that enables Large Language Models (LLMs) to generate more accurate and contextual responses, significantly improving their practicality in real-world applications (Es et al., 2024 ; Salemi and Zamani, 2024) . Compared to traditional large model Supervised Fine-Tuning (SFT) and In Context Learning (ICL) (Taori et al., 2023 ; Zhang et al., 2023) methods, RAG can adapt to  \nNaiveRAG  \n What is the full name of 'CA' airline?  \n Retrieve from Corpus  CA is an airline.  \n California Airlines is a small regional carrier.  \n Air China is a major airline in Asia.  \n Generation  \n The full name is California Airlines.  \n Wrong Answer  \nGRAG  \nWhat is the full name of 'CA' airline?  \nGeneration  \nThe full name of CA airline is Air China.  \nCorrect Answer  \nFigure 1: An example question that NaiveRAG may answer incorrectly due to inaccurate retrieval, whereas GRAG can provide the correct answer through knowledge graph relationships.  \nspecific domain knowledge to ensure that the information provided is not only relevant but also tailored to the user’s needs. In addition, RAG can dynamically update private data, so it has received widespread attention from academia and industry (Gao et al., 2023) .  \nAlthough RAG effectively leverages external and up-to-date databases, its performance heavily depends on the accuracy of the retrieved context (Zhao et al., 2024) . However, in real-world scenarios, many questions require multi-hop reasoning to answer correctly (Yang et al., 2024) . Traditional text retrieval methods, such as sparse retrieval (Robertson et al., 2009) or dense retrieval (Karpukhin et al., 2020), cannot effectively support LLMs in answering complex questions.  \nFor example, as shown in Figure 1, given a question like What is the full name of the ‘CA’airline?, traditional RAG retrieves contexts related to ‘CA’, but these contexts lack semantic connections. As a result, the LLM generates an incorrect answer. To better","cbCaitijTfsPJPe2","https://ap.wps.com/l/cbCaitijTfsPJPe2","pdf",528169,5,1,17,"English","en",105,"# Introduction\n## Retrieval-Augmented Generation and Limitations\n## Graph-Enhanced RAG and Related Work\n## Proposed NGM-RAG Framework","[{\"question\":\"Why do traditional text-based RAG methods struggle with multi-hop questions?\",\"answer\":\"Because their retrieval relies on textual similarity and often misses semantic or relational connections required to connect intermediate facts. This can lead the LLM to generate incorrect answers when the question needs multi-hop reasoning.\"},{\"question\":\"What is NGM-RAG and what components does it include?\",\"answer\":\"NGM-RAG is a graph-based retrieval-augmented generation framework that integrates graph construction, graph matching, and answer generation into a unified process. It also incorporates neural graph matching to capture relational knowledge for better retrieval and generation.\"},{\"question\":\"How does NGM-RAG perform neural graph matching?\",\"answer\":\"It combines text-based matching with Graph Neural Networks (GNNs). An adaptive weighting strategy integrates multiple matching methods and selects the most relevant contextual nodes for answer generation.\"}]",1784208991,43,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":87,"head_meta":89,"extra_data":91,"updated_unix":28},"ngm-rag-neural-graph-matching-based-retrieval-augmented-generation","",{"@graph":36,"@context":86},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/ngm-rag-neural-graph-matching-based-retrieval-augmented-generation/86159/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":24,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-07-25","2026-07-16",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"Why do traditional text-based RAG methods struggle with multi-hop questions?","Question",{"text":76,"@type":77},"Because their retrieval relies on textual similarity and often misses semantic or relational connections required to connect intermediate facts. This can lead the LLM to generate incorrect answers when the question needs multi-hop reasoning.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"What is NGM-RAG and what components does it include?",{"text":81,"@type":77},"NGM-RAG is a graph-based retrieval-augmented generation framework that integrates graph construction, graph matching, and answer generation into a unified process. It also incorporates neural graph matching to capture relational knowledge for better retrieval and generation.",{"name":83,"@type":74,"acceptedAnswer":84},"How does NGM-RAG perform neural graph matching?",{"text":85,"@type":77},"It combines text-based matching with Graph Neural Networks (GNNs). An adaptive weighting strategy integrates multiple matching methods and selects the most relevant contextual nodes for answer generation.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":93},[94,98,102,106,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":20,"slug":138},19,"General","general"]