A Survey on Knowledge-Oriented Retrieval-Augmented Generation

TL;DR

This survey advocates a knowledge-centric view of Retrieval-Augmented Generation (RAG), framing retrieval, integration, and generation as a cohesive pipeline that dynamically augments LLMs with external knowledge. It formalizes the problem as $\boldsymbol{z}=g(\boldsymbol{x})$ and $\boldsymbol{y}=f(\boldsymbol{x},\boldsymbol{z})$, and surveys basic to advanced approaches across modalities, memory, and agentic reasoning. The work catalogs fundamental challenges—knowledge selection, retrieval efficiency, and in-context reasoning—and presents a comprehensive taxonomy of methods, evaluation benchmarks, and domain applications. It highlights emerging directions such as GraphRAG, multimodal and memory-augmented RAG, and edge/trustworthy deployment, arguing that integrated, interpretable RAG systems can address real-world, knowledge-intensive tasks with greater accuracy and accountability.

Abstract

Retrieval-Augmented Generation (RAG) has gained significant attention in recent years for its potential to enhance natural language understanding and generation by combining large-scale retrieval systems with generative models. RAG leverages external knowledge sources, such as documents, databases, or structured data, to improve model performance and generate more accurate and contextually relevant outputs. This survey aims to provide a comprehensive overview of RAG by examining its fundamental components, including retrieval mechanisms, generation processes, and the integration between the two. We discuss the key characteristics of RAG, such as its ability to augment generative models with dynamic external knowledge, and the challenges associated with aligning retrieved information with generative objectives. We also present a taxonomy that categorizes RAG methods, ranging from basic retrieval-augmented approaches to more advanced models incorporating multi-modal data and reasoning capabilities. Additionally, we review the evaluation benchmarks and datasets commonly used to assess RAG systems, along with a detailed exploration of its applications in fields such as question answering, summarization, and information retrieval. Finally, we highlight emerging research directions and opportunities for improving RAG systems, such as enhanced retrieval efficiency, model interpretability, and domain-specific adaptations. This paper concludes by outlining the prospects for RAG in addressing real-world challenges and its potential to drive further advancements in natural language processing.

Figures (8)

• Figure 1: A Framework for Organizing RAG Works. The timeline spans from 2020 to the present, categorizing RAG-related research into three main areas: Basic (including RAG Learning and RAG Framework), Advanced, and Evaluation. Key milestones in language models (GPT-3, ChatGPT, GPT-4) are marked along the timeline.
• Figure 2: Overview of RAG. The framework consists of three main components: (1) A query is processed by an LLM with its internal knowledge, (2) External knowledge is retrieved based on the query, and (3) Knowledge integration combines both internal and external knowledge to generate the final answer.
• Figure 3: Fundamentals and objectives of RAG, including user intent understanding, knowledge retrieval, knowledge integration, answer generation and RAG evaluation.
• Figure 4: Basic RAG Approaches, including multi-source knowledge, embedding, indexing, retrieval and generation.
• Figure 5: Diverse knowledge utilized by RAG, including structured, semi-structured, unstructured and multimodal knowledge.