[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85827-en":3,"doc-seo-85827-105":29,"detail-sidebar-cat-0-en-105":91},{"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":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},85827,8796095461564,"Liam","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",8,"Research & Report","MC-RAG System Structure Driven RAG System for Multi Constraint Queries","Retrieval-Augmented Generation (RAG) systems struggle with complex multi-constraint questions, often producing constraint violations, factual inconsistencies, or hallucinations. The STRUCTURE-DRIVEN RAG SYSTEM FOR MULTI-CONSTRAINT QUERIES (MC-RAG) reformulates retrieval as subgraph matching over a knowledge graph. Semantic and structural embeddings combined with path-level indexing enable interpretable, structure-aware, and constraint-consistent retrieval and generation. A live demonstration supports medical or encyclopedic multi-constraint queries with constraint visualization, matching steps, and end-to-end explainable outputs.","MC-RAG System: A Structure-Driven RAG System for Multi-Constraint Queries  \nXiao Zhang 1 , Yang Wan 1 , Yi Li2 , Miao Xie 1 , Chunli Lv 1 ,3  \n1 College of Information and Electrical Engineering, China Agricultural University, China  \n2 College of Computing and Data Science, Nanyang Technological University, Singapore  \n3 Key Laboratory of Agricultural Informatization Standardization, Ministry of Agriculture and Rural  \nAffairs, China  \n[B20243080802@cau.edu.cn](B20243080802@cau.edu.cn), [SY20253082208@cau.edu.cn](SY20253082208@cau.edu.cn), [liyi0067@e.ntu.edu.sg](liyi0067@e.ntu.edu.sg),  \n[0520shui@163.com](0520shui@163.com), [lvcl@cau.edu.cn](lvcl@cau.edu.cn)  \narXiv :2607 . 10 15 1v 1 [ cs .IR] 11 Jul 2026  \nAbstract  \nRetrieval-Augmented Generation (RAG) systems are widely adopted in question answering, yet they often fail to satisfy complex multi-constraint queries, leading to constraint violations, factual inconsistencies, or hallucinations. We present STRUCTURE-DRIVEN RAG SYSTEM FOR MULTI-CONSTRAINT QUERIES(MC-RAG), a structure-driven RAG system that reformulates retrieval as a subgraph matching problem over a knowledge graph. By integrating semantic and structural embeddings with path-level indexing, MC-RAG performs interpretable, structure-aware, and constraint-consistent retrieval and generation.  \nDuring the demonstration, participants can input medical or encyclopedic multi-constraint queries, visualize how the system parses constraints, performs structural matching, and generates answers, thereby experiencing an end-to-end, interactive, and explainable RAG pipeline. A demo video is available at [https://youtu.be/J8kahzmAnu0](https://youtu.be/J8kahzmAnu0) .  \n1 Introduction  \nRetrieval-Augmented Generation (RAG) [Fan et al., 2024] has become a core framework in knowledge-driven question answering systems [Izacard and Grave, 2021; Borgeaudet al., 2022], where external documents are retrieved to ground the generation of LLMs. While effective for simple, factual queries, real-world user queries often impose multiple constraints to be satisfied simultaneously, rather than matched by loosely related texts to provide rough knowledge. For example, in the nutritional scenario shown in Fig.1, a user may ask: “Which food provides abundant protein, contains vitamin D, offers calcium beneficial for bone health, and helps with weight management?”. The answer must jointly satisfy five constraints (i.e. containing protein, calcium, and vitamin D) . However, existing RAG systems struggle to address such multi-constraint requirements. Mainstream RAG systems (i.e., chunk-based RAG systems [Lewis et al., 2020]) retrieve documents solely by semantic similarity between the query and text chunks, ignoring the constraints in queries [Karpukhin et al., 2020; Xu et al., 2024;  \nFigure 1: Multi-constraint query example.  \nKhattab and Zaharia, 2020] . Recent advancements in graphbased RAG systems [Guo et al., 2025; Liang et al., 2025; Edge et al., 2024; He et al., 2024] aim to mitigate the problem by organizing external knowledge as a knowledge graph and retrieves entities or subgraphs most similar to the query. Yet, such methods still rely on similarity ranking over local graph structures, and thus cannot ensure precise satisfaction of single or multiple constraints.  \nTo address this limitation, we propose MC-RAG, a demonstration system built upon our structure guided RAG framework [Xie et al., 2026] . It integrates constraint parsing, structural subgraph retrieval, constraint verification, and generation grounded in retrieved evidence into an interactive RAG workflow. Our core idea is to replace similarity ranking with constraint-satisfying subgraph matching over a knowledge graph [Han et al., 2013; Kim et al., 2018; Xie et al., 2018; Xie et al., 2017; Zhang et al., 2024], and use the matched subgraph as verifiable evidence to guide generation. Specifically, the system first parses the natural language query into a query graph, where the ","cbCairyRHq2Bjyjk","https://ap.wps.com/l/cbCairyRHq2Bjyjk","pdf",869667,1,5,"English","en",105,"# Introduction\n# System Architecture","[{\"question\":\"Why do existing RAG systems struggle with multi-constraint queries?\",\"answer\":\"They often retrieve passages using only semantic similarity between the query and text chunks, which ignores the specific constraints that must all be satisfied. This can lead to constraint violations, inconsistencies, or hallucinations.\"},{\"question\":\"How does MC-RAG perform retrieval to satisfy constraints?\",\"answer\":\"MC-RAG parses the natural-language query into a query graph, then retrieves knowledge-graph subgraphs that are isomorphic or approximately isomorphic to the query graph. The matched subgraph serves as verifiable evidence for generation.\"},{\"question\":\"What indexes and techniques does MC-RAG use for efficient structural retrieval?\",\"answer\":\"MC-RAG integrates semantic and structural embeddings with path-level indexing, and uses an R*-Tree index to organize external knowledge. This is designed to avoid the NP-complete bottleneck of subgraph isomorphism on large-scale knowledge graphs and improve retrieval efficiency.\"}]",1784206511,13,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"mc-rag-system-structure-driven-rag-system-for-multi-constraint-queries","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/mc-rag-system-structure-driven-rag-system-for-multi-constraint-queries/85827/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why do existing RAG systems struggle with multi-constraint queries?","Question",{"text":75,"@type":76},"They often retrieve passages using only semantic similarity between the query and text chunks, which ignores the specific constraints that must all be satisfied. This can lead to constraint violations, inconsistencies, or hallucinations.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does MC-RAG perform retrieval to satisfy constraints?",{"text":80,"@type":76},"MC-RAG parses the natural-language query into a query graph, then retrieves knowledge-graph subgraphs that are isomorphic or approximately isomorphic to the query graph. The matched subgraph serves as verifiable evidence for generation.",{"name":82,"@type":73,"acceptedAnswer":83},"What indexes and techniques does MC-RAG use for efficient structural retrieval?",{"text":84,"@type":76},"MC-RAG integrates semantic and structural embeddings with path-level indexing, and uses an R*-Tree index to organize external knowledge. This is designed to avoid the NP-complete bottleneck of subgraph isomorphism on large-scale knowledge graphs and improve retrieval efficiency.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,109,114,119,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":21,"slug":137},19,"General","general"]