Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

A Systematic Literature Review of Retrieval-Augmented Generation: Techniques, Metrics, and Challenges

Created by
  • Haebom

Author

Andrew Brown, Muhammad Roman, Barry Devereux

Outline

This paper conducted a systematic literature review of 128 highly cited research papers on Retrieval-Augmented Generation (RAG) published from 2020 to May 2025. Papers were collected from databases such as the ACM Digital Library, IEEE Xplore, Scopus, ScienceDirect, and DBLP and analyzed based on the PRISMA 2020 framework. This paper categorizes RAG datasets, architectures, and evaluation methods, and comprehensively analyzes the empirical evidence on the effectiveness and limitations of RAG to clarify current research trends, highlight methodological gaps, and suggest directions for future research priorities. To mitigate citation delay bias, a low citation threshold was applied to papers published in 2025.

Takeaways, Limitations

Takeaways:
Clearly presents the current state of RAG research.
Provides a comprehensive list of RAG's datasets, architectures, and evaluation methods.
We comprehensively analyze the empirical evidence on the effectiveness and limitations of RAG.
Provides direction and priority for future RAG research.
We propose a method to mitigate citation delay bias.
Limitations:
Limitations in the scope of the study due to searches limited to specific databases.
Possible selection bias based on citation count criteria.
Data only up to May 2025 is included, so it may not fully reflect the latest research trends.
👍