A Review of Retrieval-Augmented Generation: Architectural Evolution, Key Breakthroughs, and Vertical-Domain Practices

Main Article Content

Zihe Zhu
College of Hydraulic Science and Engineering, Northeast Agricultural University, Harbin, 150030, China

Bin Chen
College of Hydraulic Science and Engineering, Northeast Agricultural University, Harbin, 150030, China

DOI: https://doi.org/10.70267/cai.v3n4.193206

Keywords

retrieval-augmented generation, large language model, knowledge graph, multimodal fusion, vertical-domain application

Abstract

Large language models (LLMs) rely on massive pretraining based on the Transformer architecture and show significant emergent effects in general tasks. However, the nature of their statically parameterized knowledge causes vertical-domain applications to face three core bottlenecks: knowledge hallucination, delayed timeliness, and loss in long contexts. Retrieval-Augmented Generation (RAG) realizes the decoupling of computational logic and factual memory through the paradigm of “retrieve first and generate afterward”. It can dynamically inject domain knowledge without fine-tuning and has become a mainstream technical path for the implementation of trustworthy artificial intelligence (AI). This study systematically reviews the evolution of RAG technology from 2024 to 2026, from Naive RAG to Advanced RAG, Modular RAG, and graph-enhanced RAG. It analyzes in depth the breakthrough progress of four key technologies, namely hybrid retrieval, multimodal document parsing, generation control, and lightweight deployment. Combined with typical application cases in complex equipment manufacturing, energy and power, public services, and other fields, it explains the engineering value of RAG. Finally, it discusses the evaluation system, existing limitations, and future trends, providing a reference for academic research and practical implementation of RAG technology.

Abstract 20 | PDF Downloads 20

References

  • [1]Wang, H., Xiao, X., & Huang, Z. (2026). Research on the consistency governance mechanism for enterprise system interpretation based on LLM and RAG. Communications Enterprise Management, (5), 66-72.
  • [2]Ren, H. Y., Sun, Z., Liao, S., Mao, D. M., Zhang, L., & Zhang, S. W. (2026). A review of the application of large language models in network threat analysis. Journal of Computer Applications, 1–26. https://doi.org/10.11772/j.issn.1001-9081.2026010138.
  • [3]Gu, X., Qin, S., Shu, Y., Ma, B., Liu, F., & Liu, M. A lightweight cybersecurity vulnerability risk awareness method based on RAG. Computer Science, 53(S1), 863-872.
  • [4]Zhang, W., Wang, Z., Li, J., He, Y., Xu, J., & Liu, Y. (2026). RAG-Risk: A large-model hazardous chemical accident root-cause prediction method based on RAG. Journal of Systems Science and Mathematical Sciences, 1-21. https://link.cnki.net/urlid/11.2019.O1.20260526.1024.014
  • [5]Luo, J. (2026). Design and engineering application of a railway four-electricity knowledge-enhanced retrieval system based on the RAG architecture. Modern Information Technology, 10(11), 73-80. https://doi.org/10.19850/j.cnki.2096-4706.2026.11.014
  • [6]Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., ... & Kiela, D. (2020). Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in neural information processing systems, 33, 9459-9474.
  • [7]Dai, R. (2026). Research method for an MBSE data foundation based on RAG-large language models. Modern Navigation, 17(3), 191-196.
  • [8]Wang, H., Hou, Q., Fang, Z., Jiang, H., & Shen, W. (2026). Construction and application of an intelligent knowledge system for offshore oilfield logging based on improved Retrieval-Augmented Generation (RAG). Mud Logging Engineering, 37(2), 1-8.
  • [9]Xue, Z., Zhang, Y., Zheng, Z., Li, C., & Jiang, H. (2026). Intelligent power retrieval engine based on knowledge graph path reasoning and large models. Power System Technology, 50(8), 3263-3280. https://doi.org/10.13335/j.1000-3673.pst.2025.1205
  • [10]Yao, M. (2026). Progress of RAG and its application in monitoring systems. China Computer & Communication, 38(14), 133-135.
  • [11]Guo, L., Chen, J., & Li, J. (2026). Research on a hybrid general document parsing method for RAG. Software Guide, 1-7. https://link.cnki.net/urlid/42.1671.TP.20260721.1641.020
  • [12]Zhang, Y. (2026). Research on an intelligent recipe recommendation system based on Retrieval-Augmented Generation. China Computer & Communication, 38(16), 86-88.
  • [13]Feng, Y., Xue, S., Wang, S., Shu, H., Zhang, W., & Li, Q. (2026). Intelligent chunking of design documents based on layout parsing and LLM. Electric Power Survey & Design, (7), 84-88. https://doi.org/10.13500/j.dlkcsj.issn1671-9913.2026.07.015
  • [14]Xue, H., Zhang, Z., Liang, W., & Wang, J. (2026). Research on an LLM semantic communication knowledge synchronization method based on RAG retrieval enhancement. China-Arab States Science and Technology Forum (Chinese and English), (7), 81-86.
  • [15]Chi, Z., Pan, X., Huang, X., Zhou, X., & Huang, X. (2026). Optimization of a RAG system in policy consultation based on hybrid vector and tag retrieval. Industrial Control Computer, 39(7), 96-97, 100.
  • [16]Zhang, Q., Qu, P., & Zuo, L. (2026). Intelligent question-answering system based on large language models and improved RAG. Computer Systems & Applications, 1-10. https://doi.org/10.15888/j.cnki.csa.010297
  • [17]Xie, B., He, N., Zhang, R., & Yang, Y. (2026). RAG-SymbolicDet: Visual event symbolic reasoning with multi-strategy Retrieval-Augmented Generation. Computer Engineering and Applications, 1–12. https://link.cnki.net/urlid/11.2127.tp.20260727.1343.004
  • [18]Zhang, X., Chen, S., Xie, F., Hu, Y., Yang, H., Wang, L., & Xie, Q. (2026). A soil acidification obstacle diagnosis system based on Retrieval-Augmented Generation (RAG) coordinated with large models: A case study of Chengmai, Hainan. Journal of Agro-Environment Science, 1–25.
  • [19]Wang, Y., Sun, L., & Fu, Z. (2026). Research on scientific writing knowledge services integrating dual-database collaborative RAG under the AI4S paradigm: A case of AI-domain scientific writing. Journal of Modern Information, 46(8), 3-16.
  • [20]He, Q., Li, Q., Guo, H., & Chen, Q. (2026). Research on a retrieval-augmented generation method based on dynamic adaptive reinforcement learning. Journal of Frontiers of Computer Science and Technology, 1-15. https://link.cnki.net/urlid/11.5602.tp.20260727.1223.006
  • [21]Xu, G. (2026). Practice and enlightenment of intelligent question answering with large language models driven by the Coze + RAG architecture in libraries: A case study of Xiamen University’s “Law Map Assistant”. Value Engineering, 45(20), 165-168.
  • [22]Pan, X. (2026). Construction and application of an intelligent reference consultation system for university libraries based on the RAG architecture. Journal of Academic Library and Information Science, 44(4), 15-21.
  • [23]Huang, Z., & Qiao, X. (2026). Construction and application of a classroom intelligent assistant based on AIGC and RAG technology. Electronic Component and Information Technology, 10(5), 216-220. https://doi.org/10.19772/j.cnki.2096-4455.2026.05.061
  • [24]Wu, X., & Pei, C. (2026). Research on an intelligent tutoring system based on large-model RAG technology. China-Arab States Science and Technology Forum (Chinese and English), (6), 92-96.
  • [25]Lin, S. (2026). Research on a course intelligent question-answering system integrating knowledge graphs and RAG. Fujian Computer, 42(5), 68-73. https://doi.org/10.16707/j.cnki.fjpc.2026.05.012
  • [26]Liu, Y., Dong, S., Zhang, S., Cheng, Y., & Zhang, S. (2026). Design and implementation of an enhanced large language model learning system coordinating RAG and knowledge graphs: A case study of higher vocational Python and data structure courses. Computer Knowledge and Technology, 22(15), 137-140. https://doi.org/10.14004/j.cnki.ckt.2026.0697
  • [27]Wang, J., & Ma, B. (2026). Application of Graph-RAG technology based on structural semantic collaborative optimization in intelligent inquiry systems for equipment operation and maintenance. Journal of Beijing University of Chemical Technology (Natural Science Edition), 1-10. https://link.cnki.net/urlid/11.4755.TQ.20260602.1102.002
  • [28]Liu, X., Xu, L., Ma, H., & Liu, S. (2026). Research on RAG question-answering optimization for fault diagnosis of coal mine electromechanical equipment. Coal Technology, 45(6), 139-143. https://doi.org/10.13301/j.cnki.ct.2026.06.026
  • [29]Xiong, L., Du, S., Shi, H., Hu, J., Yang, Y., & Li, T. (2026). Government hotline question-answering system based on a knowledge-enhanced large language model architecture. Journal of Computer Applications, 46(6), 1721-1727.
  • [30]Wang, Y., Liu, D., Che, S., & Wei, Y. (2026). Technical research on a grass-roots grid question-answering assistant based on large models and RAG. Jiangsu Communications, 42(3), 19-24, 50.
  • [31]Zhao, J. (2026). Research on a BIM knowledge graph intelligent retrieval system based on Graph-RAG. Modern Information Technology, 10(11), 62-67. https://doi.org/10.19850/j.cnki.2096-4706.2026.11.012
  • [32]Yang, Y., Yang, C., Wu, H., Zhu, G., Liu, M., & Ren, Z. (2026). Hierarchical criterion-driven profile construction and graph retrieval-augmented bidding screening method. Journal of Frontiers of Computer Science and Technology, 1–17.
  • [33]Zhang, Y., Liu, Y., Li, L., Qi, D., Xu, D., Li, Y., & Wei, H. (2026). A ship RAG system enhanced by ship professional vocabulary. Journal of Shanghai Ship and Shipping Research Institute, 49(3), 33-39, 62.
  • [34]Sun, Y., Wang, Z., Liu, H., Wang, W., & Jiang, M. (2026). Construction of an intelligent question-answering system for Python programming based on LLM and Advanced RAG. Research and Exploration in Laboratory, 45(8), 104-110, 116. https://doi.org/10.19927/j.cnki.syyt.2026.08.015
  • [35]Li, Z., Jiang, D., & Wang, H. (2026). Exploration and design of a university knowledge question-answering system based on large models and RAG technology. China Computer & Communication, 38(12), 41-43.
  • [36]Wang, C., Wang, H., Zhang, B., Zhou, M., & Xu, Y. (2026). Research and practice of corpus governance for an urban rail quality and safety large model driven by RAG. Tunnel Construction, 1-11. https://link.cnki.net/urlid/44.1745.U.20260727.1033.002
  • [37]Gao, J., Ai, X., & Zhang, S. (2026). LLM aquatic nutrition recommendation architecture based on Hybrid RAG. Journal of Shanghai Ocean University, 35(4), 855-866.
  • [38]Dou, J., Wang, X., Qiao, J., Jin, M., Li, J., Li, Y., & Han, Y. (2026). Distribution network maintenance intelligent auxiliary decision-making framework based on knowledge graph Retrieval-Augmented Generation and large language models. Advanced Engineering Sciences, 58(4), 67-79.
  • [39]Shen, J., Chen, H., & Chen, X. (2026). Multi-RAG: A distributed Retrieval-Augmented Generation framework for cross-domain data. Computer Science, 53(S1), 472-478.
  • [40]Xu, Y., Ni, H., Jiang, M., & Mei, Z. (2026). Research on an intelligent auxiliary design method for ship pipeline systems based on RAG and knowledge graphs. Ship Materials & Market, 34(6), 16-19. https://doi.org/10.19727/j.cnki.cbwzysc.2026.06.006
  • [41]Zhang, L., Huang, C., Ding, W., Pan, Y., Zhu, Z., & Yang, H. (2026). Intelligent transformation of misinformation detection: Technical system, dilemmas, and prospects based on large language models. Information Security and Communications Privacy, (7), 9-25.
  • [42]Wang, Y., Dai, S., Li, J., Wang, G., & Sun, J. (2026). A review of the application of Retrieval-Augmented Generation in misinformation detection. Computer Engineering and Applications, 1-33. https://link.cnki.net/urlid/11.2127.TP.20260804.1915.006
  • [43]Jing, Z., Hu, X., & Hu, J. (2026). Generation of community security risk disposal recommendations based on hierarchical graph retrieval augmentation. Journal of Computer Applications, 1-13. https://link.cnki.net/urlid/51.1307.TP.20260723.1744.004
  • [44]Feng, Z., Zhang, Y., Yang, X., & Chen, T. (2026). Theoretical construction and framework implementation of trusted smart data for patent literature. Library Tribune, 1-16. https://link.cnki.net/urlid/44.1306.G2.20260720.1026.002
  • [45]Ren, Z., Li, Y., Ma, Y., Chen, Q., Li, J., Yang, S., & Liu, L. (2026). Efficacy evaluation of domestic lightweight large language models integrated with Retrieval-Augmented Generation for lung cancer specialty question answering. Chinese Journal of Clinical Thoracic and Cardiovascular Surgery, 33(9), 1403-1411.
  • [46]Li, Y., Li, D., Meng, X., Liu, X., & Cui, H. (2026). Research on localized detection instruments based on AI agents. Automation & Instrumentation, 41(7), 106-109, 114. https://doi.org/10.19557/j.cnki.1001-9944.2026.07.023
  • [47]Tang, L., Li, W., Huang, P., Huang, X., Wang, Q., & Liu, J. (2026). Intelligent recommendation system for Chinese patent medicine based on cloud services and RAG technology. Computer Science, 53(S1), 213-219.
  • [48]Yue, C. (2026). Research on a military equipment domain knowledge question-answering system integrating knowledge graphs and large language models [Doctoral dissertation, Yantai University]. https://doi.org/10.27437/d.cnki.gytdu.2026.000149
  • [49]Sun, C., Lan, T., Wu, Z., Cheng, D., Shi, X., & Jiang, S. (2026). Travel route schematic generation framework based on LLM and RAG. Surveying and Mapping, 49(3), 169-174.
  • [50]Li, M., Zhang, C., Hu, H., Huang, L., & Zhang, Z. (2026). Auxiliary application of large language models in finite element modeling of ship structures. Journal of Shanghai Ship and Shipping Research Institute, 49(3), 27-32.
  • [51]Li, W., Zhang, Z., Tao, W., Liu, C., Wang, L., & Li, R. (2026). Intelligent frequency regulation of power systems assisted by large language models. High Voltage Engineering, 52(7), 3010-3021. https://doi.org/10.13336/j.1003-6520.hve.20250201
  • [52]Liu, J., Zhou, Q., Guo, R., & Feng, J. (2026). Applied research based on large language models and Retrieval-Augmented Generation: A case study of a feasibility study report consultation system. China Engineering Consulting, (7), 97-105.
  • [53]Zhao, Y., Huang, W., & Liu, Y. (2026). Construction of a smart city emergency intelligence service agent integrating large models and Hybrid-RAG. Journal of Modern Information, 1-14. https://link.cnki.net/urlid/22.1182.G3.20260706.1521.004
  • [54]Feng, S., & Yang, H. (2026). Construction and empirical study of a Retrieval-Augmented Generation evaluation benchmark based on South China Sea news dynamics. Journal of Intelligence, 1-8. https://link.cnki.net/urlid/61.1167.g3.20260728.1418.010

Article Sidebar

Full Text
PDF
Information
Copy Publication Information
Published
Oct 30, 2026
Issue
Vol. 3 No. 4 (2026)
Section
Research Articles
Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.

How to Cite

Zhu, Z., & Chen, B. (2026). A Review of Retrieval-Augmented Generation: Architectural Evolution, Key Breakthroughs, and Vertical-Domain Practices. Computers and Artificial Intelligence, 3(4), 193-206. https://doi.org/10.70267/cai.v3n4.193206