[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-81739-en":3,"doc-seo-81739-105":30,"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":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},81739,4810365810221,"Aurora","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",8,"Research & Report","AGE: Adaptive-masking for Graph Embedding in Graph Retrieval-Augmented Generation","GraphRAG extends retrieval-augmented generation (RAG) by using graph-structured data as external knowledge for large language models, but graph-to-text latent feature misalignment often limits frozen LLM performance. AGE introduces an adaptive-masking graph embedding method using a Transformer-based, mask-based self-supervised learning scheme. Architecture is designed to mirror text embedding encoders while avoiding inefficiency from masking key nodes. AGE predicts non-key nodes via a learnable, reinforcement-learning node sampler, improving GraphQA accuracy across four benchmarks.","arXiv :2607 .00052v 1 [ cs .IR] 30 Jun 2026  \nAGE: Adaptive-masking for Graph Embedding in Graph Retrieval-Augmented Generation  \nBao Long Nguyen Huu 1 and Atsushi Hashimoto2  \n1 OMRON Corporation  \n2 OMRON SINIC X Corporation  \nAbstract. GraphRAG is an extension of retrieval-augmented generation (RAG) that supports large language models (LLMs) by referring to graph-structured data as external knowledge. While this technique ideally captures intricate relationships, it often struggles with graph representations for LLMs, particularly for frozen LLMs, due to the misalignment between graph-based and text-based latent features. We tackle this issue by introducing the Adaptive-masking for Graph Embedding (AGE) .  \nAGE employs a Transformer in a mask-based self-supervised learning (SSL) approach. We designed the architecture similar to text embedding encoders, addressing the latent feature misalignment. In contrast to natural language texts, graphs are concise representations, and there exist key nodes that hold dominant contextual information, which are challenging to predict from their surroundings. Masking such key nodes leads to inefficiency in the SSL process. Therefore, AGE focuses on predicting nodes apart from key nodes, utilizing a learnable node sampler. Our experimental results indicate that AGE significantly improves approaches using non-parametric search component in GraphQA tasks, achieving superior accuracy across four benchmark datasets with distinct characteristics.  \nKeywords: Knowledge Graph Question Answering · Retrieval-Augmented Generation · Reinforcement Learning · Self Supervised Learning.  \n1 Introduction  \nLarge Language Models (LLMs) such as GPT [58, 59], Claude [2], Gemini [71], Qwen [83], and LLaMA [24] have significantly advanced natural language understanding and generation capabilities. Retriever-Augmented Generation (RAG)[20,23,80] integrates query-relevant information into the generation process, enabling LLMs to access and utilize domain-specific knowledge beyond their pretraining corpus. However, although RAG enhances LLMs with external data, it may struggle to capture essential structured relationships, reducing search precision and reasoning effectiveness [84,89] . Graph Retriever Augment Generation (GraphRAG) [18,55] is a technology that uses graphs to overcome the limitations of RAG. Graph data, represented by nodes (entities) and edges (relationships), clearly presents complex relationships. This provides several benefits, such as  \n2 F. Author et al.  \nSelf Attention Layers of  \nLLM   \nEmbedding Inference  \nGraph Encoder  \nOutput  \n “No”  \nFig. 1: Overview of GraphRAG with the proposed Adaptive-masking for Graph Embedding (AGE) embedding. 1) Retrieval: Find graph elements relevant to the query using a non-parametric process. 2) Subgraph Construction: Extend retrieved graph elements with their adjacencies [32] . 3) Embedding: Use tokenizer and text embedder for textualized graph and query. Apply AGE for structured relationships of the graph.  \n4) Inference: Input embeddings into LLM to generate an answer.  \nfacilitating data integration [36], improving search accuracy [9, 28], enhancing inference capabilities [27, 29], and reducing hallucinations [32] . By capturing sub-graphs, the broader context and interconnections within the graph structure can be captured, enabling comprehensive information to be accessed for LLMs enhance the performance in domain-specific tasks.  \nThis study investigates GraphRAG methods that operate within practical computational costs. Fine-tuning LLMs can enhance GraphRAG performance, yet it is resource-intensive. Instead, previous methods often focused on the retrieval module as it is a key factor for GraphRAG performance. Trainable retrievers, such as LLM-based retrievers [50,66] realize a higher retrieval accuracy. However, this strategy still requires significant computational overhead. Non-parametric retrievers [32,85] are efficient and low-cost but may contain","cbCaitAEZ4zgoruO","https://ap.wps.com/l/cbCaitAEZ4zgoruO","pdf",4505601,4,1,38,"English","en",105,"# Introduction\n## GraphRAG and the challenge of structured relationships\n## Practical constraints: frozen LLMs and non-parametric retrievers\n## Proposed method: Adaptive-masking for Graph Embedding (AGE)\n## Experimental impact on GraphQA","[{\"question\":\"What problem does AGE address in GraphRAG with frozen LLMs?\",\"answer\":\"AGE targets misalignment between graph-based latent features and the LLM’s text-based embedding space, which can prevent effective structured reasoning when the LLM is frozen.\"},{\"question\":\"How does AGE implement its training approach?\",\"answer\":\"AGE uses a Transformer with a mask-based self-supervised learning approach and designs the architecture to resemble text embedding encoders to reduce latent feature misalignment.\"},{\"question\":\"Why does AGE avoid masking key nodes during self-supervised learning?\",\"answer\":\"The document explains that masking dominant key nodes can make SSL inefficient because their contextual information is difficult to predict from surrounding context.\"}]",1784175754,96,{"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":86,"head_meta":88,"extra_data":90,"updated_unix":28},"age-adaptive-masking-for-graph-embedding-in-graph-retrieval-augmented-generation","",{"@graph":36,"@context":85},[37,53,68],{"@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":20},"https://docshare.wps.com/document/age-adaptive-masking-for-graph-embedding-in-graph-retrieval-augmented-generation/81739/",{"url":52,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-24","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},"What problem does AGE address in GraphRAG with frozen LLMs?","Question",{"text":75,"@type":76},"AGE targets misalignment between graph-based latent features and the LLM’s text-based embedding space, which can prevent effective structured reasoning when the LLM is frozen.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does AGE implement its training approach?",{"text":80,"@type":76},"AGE uses a Transformer with a mask-based self-supervised learning approach and designs the architecture to resemble text embedding encoders to reduce latent feature misalignment.",{"name":82,"@type":73,"acceptedAnswer":83},"Why does AGE avoid masking key nodes during self-supervised learning?",{"text":84,"@type":76},"The document explains that masking dominant key nodes can make SSL inefficient because their contextual information is difficult to predict from surrounding context.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"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":106,"slug":138},19,"General","general"]