Lius: Translation Model Based Instructional Lingustic Using Continual Instruction Tuning In Kupang Malay

2026-06-10

AI Summary

This study addresses the significant performance degradation of large language models in translating low-resource languages—such as Kupang Malay—due to the scarcity of parallel corpora. To mitigate reliance on extensive parallel data, the authors propose a novel approach that integrates bilingual dictionary–driven explicit lexical and semantic feature instructions with continual instruction tuning (CIT), marking the first fusion of these two strategies. The resulting model, Lius, achieves consistent improvements of 4–6 points over standard instruction tuning across multiple automatic evaluation metrics and outperforms both state-of-the-art neural machine translation systems and multilingual large language models by margins of 10–13 points, substantially enhancing translation quality for low-resource languages.

Abstract

Large Language Models (LLMs) offer new potential for translation tasks but often experience performance degradation when handling low-resource languages. To address this limitation, we propose an approach for fine-tuning LLMs on a low-resource language, Kupang Malay. Our approach involves designing a set of instructions by leveraging explicit lexical and semantic features from a bilingual dictionary, and introducing Continual Instruction Tuning (CIT), a training paradigm that enables iterative instruction-based training. Experimental results demonstrate that our model, named Lius, yields notable improvements over standard instruction-tuned models by outperforming 4-6 points, and surpassing both Neural Machine Translation (NMT) and Multilingual LLM models by 10-13 points on several evaluation metrics. These findings highlight the potential of our approach to mitigate the reliance on large-scale parallel data in low-resource language translation.

Problem

• low-resource languages
• machine translation
• Large Language Models
• Kupang Malay
• parallel data scarcity

Innovation

• Continual Instruction Tuning
• Low-resource Language Translation
• Instruction-based Fine-tuning
• Kupang Malay
• Bilingual Dictionary Integration