Retrieval-Augmented Large Language Models for Robust Context-Aware Natural Language Understanding
International Journal of Innovative Science and Research Technology · 07.04.2026 · 3420
DOI: 10.38124/ijisrt/26mar1756 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Innovative Science and Research Technology, online erschienen am 07.04.2026. Verfasst von Utsha Sarker und 5 weiteren Autor:innen. Die Arbeit nennt 25 Literaturangaben.
Abstract (Originalsprache)
Large Language Models (LLMs) have been shown to have remarkable capabilities in natural language understanding; however, they still have some limitations such as the outdated knowledge, the lack of domain-specific awareness and the hallucination of incorrect information. These problems are induced by the fact that LLMs are mainly based on parametric knowledge stored during the training process, that is not dynamically updated and verified . To combat such challenges, this paper introduces an improved Retrieval-Augmented Generation (RAG) to address these underlying challenges which combines an improved context aware retrieval mechanism with the gating based prompt augmentation strategy. The proposed approach selectively filters and ranks the retrieved documents based on context-awareness gate before injecting them to the LLM, which would improve the relevance and reduce the noise in the generated responses.In the paper we validate the proposed method using benchmark data such as SQuAD, domain-specific question answering data sets as well as dialogue data sets where we compare with baseline models such as vanilla LLMs and standard RAG pipelines. Experimental results show that our method can provide much better results in terms of Exact Match (EM), F1-score and Fact consistency compared to traditional methods. These findings are consistent with recent studies showing the value of RAG in enhancing factual grounding and reducing hallucinations in LLMs 1. Contributions: In this paper, we propose a novel context-aware RAG architecture, which provides a retrieval filtering mechanism. Following the review, we design an improved prompt integration strategy for improved knowledge grounding. We empirically show better performance on several NLP benchmarks.
Zitieren
APA
Sarker, U., Biswas, A., Ali, I., Vaishnav, L., ., H., & Agarwal, P. (2026). Retrieval-Augmented Large Language Models for Robust Context-Aware Natural Language Understanding. International Journal of Innovative Science and Research Technology, 3420. https://doi.org/10.38124/ijisrt/26mar1756
Vancouver
Sarker U, Biswas A, Ali I, Vaishnav L, . H, Agarwal P. Retrieval-Augmented Large Language Models for Robust Context-Aware Natural Language Understanding. International Journal of Innovative Science and Research Technology. 2026:3420. doi:10.38124/ijisrt/26mar1756
BibTeX
@article{Sarker2026,
author = {Sarker, Utsha and Biswas, Archy and Ali, Ikram and Vaishnav, Lalit and ., Harsh and Agarwal, Priyanshu},
title = {Retrieval-Augmented Large Language Models for Robust Context-Aware Natural Language Understanding},
journal = {International Journal of Innovative Science and Research Technology},
year = {2026},
pages = {3420},
doi = {10.38124/ijisrt/26mar1756}
}
EUVIMED