Home / Current Issue / Paper 1717257
Enhancing Retrieval-Augmented Generation for Question Answering using Hybrid Retrieval and Re-ranking Techniques
Subject area: Science,Engineering and Technology · Area of research: Genrative AI
DOI: https://doi.org/10.64388/IREV9I11-1717257
Abstract
Large Language Models (LLMs) have significantly improved question-answering systems but often suffer from hallucination and reliance on static knowledge. Retrieval-Augmented Generation (RAG) addresses these limitations by incorporating external knowledge; however, its performance largely depends on the quality of retrieved context. This paper proposes an enhanced RAG-based question answering system that integrates hybrid retrieval and re-ranking techniques to improve answer accuracy. The system combines dense and sparse retrieval methods using Reciprocal Rank Fusion (RRF) to improve document relevance, followed by a cross-encoder-based re-ranking module for refined context selection. A vector database is used for efficient semantic search, and a large language model generates context-aware responses. The proposed approach is evaluated on the TriviaQA dataset using Exact Match (EM), F1 Score, BERTScore, and Knowledge Gap Detection (KGD). Experimental results show that the system achieves an Exact Match score of 85%, outperforming baseline RAG approaches such as QA-RAG. The results demonstrate that hybrid retrieval and re-ranking significantly enhance the accuracy and reliability of RAG-based question answering systems.
Keywords
Retrieval-Augmented Generation (RAG), Question Answering, Large Language Models (LLMs), Natural Language Processing (NLP)
How to cite this paper
@article{1717257,
author = {Nemalipuri Bala Thirumalesh, Nakkanaboina Bhavya Sri, Kedasu Hema Sri},
title = {Enhancing Retrieval-Augmented Generation for Question Answering using Hybrid Retrieval and Re-ranking Techniques},
journal = {Iconic Research And Engineering Journals},
year = {2026},
volume = {9},
number = {11},
pages = {1-5},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1717257.pdf},
abstract = {Large Language Models (LLMs) have significantly improved question-answering systems but often suffer from hallucination and reliance on static knowledge. Retrieval-Augmented Generation (RAG) addresses these limitations by incorporating external knowledge; however, its performance largely depends on the quality of retrieved context. This paper proposes an enhanced RAG-based question answering system that integrates hybrid retrieval and re-ranking techniques to improve answer accuracy. The system combines dense and sparse retrieval methods using Reciprocal Rank Fusion (RRF) to improve document relevance, followed by a cross-encoder-based re-ranking module for refined context selection. A vector database is used for efficient semantic search, and a large language model generates context-aware responses. The proposed approach is evaluated on the TriviaQA dataset using Exact Match (EM), F1 Score, BERTScore, and Knowledge Gap Detection (KGD). Experimental results show that the system achieves an Exact Match score of 85%, outperforming baseline RAG approaches such as QA-RAG. The results demonstrate that hybrid retrieval and re-ranking significantly enhance the accuracy and reliability of RAG-based question answering systems.},
keywords = {Retrieval-Augmented Generation (RAG), Question Answering, Large Language Models (LLMs), Natural Language Processing (NLP)},
month = {May},
doi = {https://doi.org/10.64388/IREV9I11-1717257}
}