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{202312,
        author = {Sangh Priya Gautam},
        title = {A Hallucination Aware Retrieval Augmented Generation Framework Using Cross Encoder Reranking and NLI Based Verification},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {12},
        pages = {6149-6152},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=202312},
        abstract = {While Large Language Models (LLMs) can perform well in many natural language processing (NLP) tasks, they frequently produce coherent yet inaccurate answers, referred to as "hallucinations. Although Large Language Models (LLMs) can perform well on many natural language processing (NLP) tasks, they can sometimes produce factually incorrect answers with high fluency, which are known as "hallucinations. These can lower reliability particularly in knowledge-rich and safety-sensitive applications. Retrieval-Augmented Generation (RAG) is a method designed to limit hallucination by grounding the generation in external information, which requires high retrieval accuracy and fact checking.
This paper proposes a novel hallucination-aware RAG framework which integrates reduction and detection at a single step of a fully automated pipeline. The system architecture combines a well-tuned SBERT-based dense retriever, a cross-encoder reranker for semantic relevance, and a Natural Language Inference (NLI) verifier to ensure logical consistency between generated responses and retrieved evidence, along with hard-negative mining. This NLI-based Validation allows automatic hallucination detection without human intervention. Through experimental results, the framework is shown to be effective in practical and scalable RAG applications with improved factual accuracy and reduced hallucination rates, while maintaining response quality.},
        keywords = {Retrieval-Augmented Generation; Hallucination Detection; Cross-Encoder Reranking; Natural Language Inference; Large Language Models},
        month = {May},
        }