Skip to content
Conference Open access

A Scalable Multilingual Query Response Model Using LLaMA 3.2-Based Language Translation and Agentic Retrieval-Augmented Generation

Sep 2026 · 2026 International Conference on Secure IoT and Cybersecurity (ICOSICS) · pp. 1614-1624 · 0 citations · 26 references

Abstract

Modern Question-Answering (QA) systems rely on Large Language Models (LLMs), the inference latency of which can vary substantially with input complexity. Although previous work has explored end-to-end throughput and model size tradeoffs, little attention has been paid to the precise relationship between the linguistic complexity of a query and the time spent in prompt evaluation. Despite increasing research attention to text ranking, most studies focus on monolingual scenarios, with a particular emphasis on English-language contexts. Improving reasoning performance in lower-resource languages is challenging, as we lack large amounts of data. Medium- and low-resource languages have been neglected despite these advancements, since the majority of attention has been on English and other high-resource languages. Therefore, a multilingual query response system is designed to bridge linguistic diversity by enabling seamless interaction across regional languages. Various regional texts are collected from benchmark datasets, as the proposed system accepts user queries in regional texts. Then, an LLaMA 3.2-based translation module is utilized to convert them into English. Following this, an Agentic Retrieval-Augmented Generation (A-RAG) model is employed to interpret the translated query and generate accurate responses. Once the response is generated in English, it is translated back into the user's original regional language using the LLaMA 3.2 model to ensure accessibility. The proposed model is useful for different applications like customer support, education, and knowledge access systems. The proposed approach enhances cross-lingual information retrieval and reduces language barriers. The effectiveness of the suggested approach in maintaining semantic accuracy is validated among conventional models.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.