MUMBAI, India, June 22 -- Intellectual Property India has published a patent application (202641048607 A) filed by Dr. Neelam Sanjeev Kumar; Harshat B; Viveks Nair; and Sreenidhi T on April 16, 2026, for Implementation Of A Rag Pipeline Using Multi-Agent Architecture.
Inventors include Dr. Neelam Sanjeev Kumar; Harshat B; Viveks Nair; and Sreenidhi T.
The application for the patent was published on June 12, 2026, under issue no. 24/2026.
Abstract: The present invention proposes an advanced Retrieval-Augmented Generation (RAG) system enhanced with a m ulti-agent architecture to improve the precision, interpretability, and reliability of large language model (LLM) outputs. Traditional RAG systems rely on a single monolithic pipeline, often leading to issues such as irrelevant retrieval, hallucination, and lack of transparency. The proposed system decomposes the RAG pipeline into multiple specialized agents, including a query understanding agent, query rewriting agent, retrieval agent, reranking agent, answer generation agent, and verification agent. Each agent is responsible for a specific task, enabling modularity, better control, and improved performance. The retrieval component is trained using dense embeddings generated from domain-specific documents and indexed using vector databases such as FAISS. A proprietary LLM accessed via API is used for query refinement and answer generation. The system is implemented with a user interface built using Streamlit. Experimental evaluation demonstrates that the multi-agent RAG architecture significantly reduces hallucination and improves answer relevance compared to traditional single-agent RAG systems.
Disclaimer: Curated by HT Syndication.
