Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{206976,
        author = {Vishal Patil and Dr. R. R. Keole},
        title = {A Review on Knowledge Graph-Based Retrieval-Augmented Generation for Reliable Scientific Response Generation},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {2},
        pages = {3816-3822},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=206976},
        abstract = {The world is rapidly evolving around AI and generative AI. From chat bots that write essays to models that sketch artworks in seconds, we are now living in an age where Large Language Models (LLMs) can read, summarize, and even draft scientific papers in a matter of minutes. Yet, when it comes to science, these LLMs still suffer from major reliability problems: they often produce unsupported claims, omit proper citations, invent references, mix up metadata, and create semantic inconsistencies between the answer and the cited papers. Recent developments such as citation aware generation, GraphRAG, claim verification tools, knowledge graph builders, and RAG evaluation frameworks have started to tackle these issues – but most solutions focus on only one aspect (e.g., improving retrieval, tightening citation integration, detecting hallucinations, or verifying individual claims). What we need is a completely new approach that moves us from a simple “retrieve then generate” workflow to a “verify first” paradigm. This review paper discusses recent developments aimed at improving the trustworthiness of LLM-based scientific writing. It examines RAG, citation-aware generation, GraphRAG, scientific claim verification, knowledge graph construction, and RAG evaluation methods. The review identifies several research gaps, including fragmented solutions that separately address retrieval, citation, hallucination, and verification. Existing studies also provide limited support for claim-level graph traversal, citation reliability assessment, contradiction detection, and auditable reasoning. Therefore, a unified verification framework is required to validate evidence and semantic consistency before generating the final scientific response.},
        keywords = {Retrieval Augmented Generation, GraphRAG, Citation Reliability, Response Verification, Claim Verification, Evidence Path Verification, Scientific LLM Systems},
        month = {July},
        }