2026· International journal of research and innovation in social science· Vol 10, pp. 13370-13379· 0 citations
TL;DR
The evidence indicates that no single RAG or vector-database configuration dominates across retrieval quality, faithfulness, latency, throughput, storage, cost, and scalability, and the review positions RAG–vector database integration as a joint retrieval-and-systems optimization problem rather than a database-selection problem alone.
Abstract
Large Language Models (LLMs) have demonstrated strong natural-language generation capability, but knowledge-intensive use remains limited by static parametric knowledge, hallucination, restricted access to proprietary information, and weak evidence traceability. Retrieval-Augmented Generation (RAG) addresses these limitations by retrieving external evidence at inference time, while vector databases provide the storage, indexing, filtering, and similarity-search mechanisms needed to operationalize retrieval at scale. This systematic literature review examines the joint design of RAG pipelines and vector database infrastructure, with emphasis on retrieval architectures, embedding and chunking choices, approximate-nearest-neighbour indexes, reranking, application domains, and end-to-end evaluation. The review was conducted using established systematic-review guidance and reported in accordance with PRISMA 2020. To strengthen analytical depth, the studies were appraised using a structured methodological-quality rubric and synthesized through a multidimensional comparison matrix. The evidence indicates that no single RAG or vector-database configuration dominates across retrieval quality, faithfulness, latency, throughput, storage, cost, and scalability. Building on these findings, this review proposes a unified evaluation framework covering retrieval effectiveness, generation quality, system efficiency, and dynamic-knowledge robustness. Persistent challenges include stale embeddings, index-update cost, knowledge freshness, security, privacy, explainability, and inconsistent benchmarking. The review therefore positions RAG–vector database integration as a joint retrieval-and-systems optimization problem rather than a database-selection problem alone.
Large language models (LLMs) generate fluent text but remain vulnerable to hallucinations, outdated knowledge, weak provenance, and costly model updates in knowledge‐intensive settings. Retrieval‐augmented generation (RAG) addresses these limitations by grounding generation in external evidence retrieved at inference time. This Advanced Review presents a PRISMA‐guided synthesis of 91 studies published between January 2020 and December 2025. It traces the evolution of RAG from early retrieve‐then‐generate pipelines to a broader design space that includes sparse, dense, and hybrid retrieval; multi‐stage reranking; evidence fusion; retrieval‐aware pre‐training; non‐parametric memory; iterative and agentic workflows; GraphRAG; multimodal RAG; multilingual retrieval; and privacy‐preserving deployment. The review examines how retrieval granularity, indexing, reranking, context construction, and retriever‐generator integration affect recall, precision, latency, robustness, and governance. It also compares explicit retrieval with long‐context prompting, vector‐based retrieval with graph‐based retrieval, and fixed pipelines with adaptive or agentic architectures. Domain requirements shape these choices: healthcare systems require evidence quality and privacy protection; legal and policy applications prioritize authoritative sources and citation fidelity; enterprise systems require freshness and access control; and educational systems must combine factual grounding with pedagogically appropriate explanations. Evaluation remains a central challenge. End‐task accuracy alone does not capture retrieval quality, faithfulness, attribution, calibration, abstention, efficiency, or resilience to prompt injection and knowledge‐base poisoning. The review therefore synthesizes current evaluation practices, promotes reproducibility‐oriented reporting through RAG Cards, and identifies open research directions for secure, efficient, transparent, and domain‐aware RAG systems.
Arman Rasool Faridi, Faraz Masood, Keshvi et al.· WIREs Data Mining and Knowle...· 0 citations
Scientific question answering requires a retrieval system to solve two distinct problems: identifying which papers are relevant and locating the supporting evidence within those papers. Conventional retrieval-augmented generation typically addresses both through similarity search over fixed-length passages, flattening document structure and separating scientific claims from their methodological and argumentative context. We present VecTree-RAG, an agentic framework that assigns these tasks to complementary retrieval mechanisms. Vector search ranks compact document and section representations across the corpus, whereas reasoning-guided traversal of source-verified section trees localizes evidence within shortlisted papers. Full text is retained in a page store and exposed progressively only after structural localization. We evaluate VecTree-RAG on 300 QASPER questions, an open-access subset of 54 LitQA2 questions, and 49 multi-document MOSAIC questions. Compared with Dense RAG, reranked Dense RAG, RAPTOR, and Search-o1, VecTree-RAG obtained the highest observed answer score on all three benchmarks, reaching 0.800 LLM-judge correctness on QASPER, 0.925 accuracy on LitQA2, and a 0.547 composite score on MOSAIC. On QASPER, its evidence-page precision was 0.274, compared with 0.046--0.071 for the baselines. LitQA2 ablations further showed that the complete vector--tree architecture required fewer inference tokens than variants without tree navigation or corpus-level vector routing. These results indicate that vector retrieval narrows the corpus-level search space and tree navigation concentrates reading on structurally relevant evidence. Although multi-turn inference remains more expensive than single-call retrieval, VecTree-RAG provides a structure-aware and traceable architecture for scientific literature question answering.
Xinyan Zhong, Yuwei Shi, Yu-Qi Wei et al.· 0 citations
The deployment of Large Language Models (LLMs) for low-resource languages is challenging due to the lack of linguistic resources, sparse digital content and the absence of structured knowledge bases. In this paper, we present an adaptive knowledge-augmented framework for Mizo Large Language Models by combining Retrieval-Augmented Generation (RAG) with continual learning. This methodology harnesses semantic retrieval with dense embeddings and FAISS indexing, adaptive evidence re-ranking, parameter-efficient fine-tuning, and incremental knowledge updating to enhance factual accuracy and decrease hallucinations. Experimental evaluation shows better retrieval performance, greater text creation quality, and superior human evaluation scores than typical multilingual LLMs and static RAG methods. Moreover, the continual learning technique allows for effective integration of newly accessible Mizo resources, without re-training the model from scratch. The suggested architecture offers a scalable, stable and reusable method for the development of intelligent language technologies for Mizo and other low-resource languages.
Vanlalropuia Ralte, Abhisake Sinha· International Journal For Mu...· 0 citations
Results demonstrate a 157% improvement in retrieval quality over a Naive-RAG baseline, with only 50 ms additional la tency, while Qwen2-VL-2B-Instruct achieved results comparable to cloud-based models in BERTScore, validate that open-source optimized SLMs, paired with advanced retrieval strategies, can provide competitive performance for document understanding without relying on cloud-based models.
Experimental results show that SAC-RAG reduces token consumption by 38%–58% at the cost of only a 1–2 percentage point EM drop, with EM actually improving after compression for reasoning-type questions, achieving the optimal quality–efficiency trade-off in terms of token consumption.
Deyu Zhang, Hongqiang Yu, Jinze Huo et al.· IEEE Access· 0 citations
Artificial Intelligence (AI) and Large Language Models (LLMs) have significantly transformed knowledge management by enabling intelligent, context-aware, and automated information access. However, standalone LLMs often suffer from limitations such as outdated knowledge, hallucinated responses, lack of domain-specific expertise, and limited transparency, reducing their reliability in enterprise and research applications. Retrieval-Augmented Generation (RAG) has emerged as an effective solution by combining language models with external knowledge retrieval, allowing responses to be generated using up-to-date and relevant information. This study proposes a comprehensive Retrieval-Augmented Generation framework for intelligent knowledge management systems. The framework integrates document acquisition, preprocessing, semantic embedding generation, vector database indexing, document retrieval, prompt augmentation, LLM-based response generation, response validation, and continuous knowledge base updates. It supports diverse knowledge sources, including enterprise databases, technical documents, digital libraries, and research repositories, while incorporating sparse, dense, hybrid retrieval, and neural reranking techniques to improve retrieval accuracy. The proposed framework is evaluated using retrieval precision, recall, F1-score, response relevance, latency, grounding accuracy, and user satisfaction. Results demonstrate improved semantic understanding, reduced hallucinations, enhanced factual correctness, and real-time knowledge updates compared with conventional keyword-based knowledge management systems. The study also discusses future directions, including multimodal RAG, graph-enhanced retrieval, federated knowledge management, continual learning, and autonomous enterprise knowledge assistants, establishing RAG as a robust foundation for trustworthy and intelligent knowledge-driven AI systems.
Louis Pouzin, J. Arsac· International Journal of Mod...· 0 citations
