Adapting Language-Specific LLMs to a Reasoning Model in One Day via Model Merging -- An Open Recipe

TL;DR

This paper presents a method to quickly adapt language-specific LLMs, like Thai, to advanced reasoning models such as DeepSeek R1 within a day, using model merging and data selection, without losing language-specific performance.

Contribution

It introduces a simple, open recipe for merging models to enhance reasoning in low-resource language LLMs efficiently and cost-effectively.

Findings

Enhanced reasoning capabilities in Thai LLM to match DeepSeek R1

Achieved this with only publicly available data and $120 computational budget

Maintained language-specific performance after merging

Abstract

This paper investigates data selection and model merging methodologies aimed at incorporating advanced reasoning capabilities such as those of DeepSeek R1 into language-specific large language models (LLMs), with a particular focus on the Thai LLM. Our goal is to enhance the reasoning capabilities of language-specific LLMs while maintaining their target language abilities. DeepSeek R1 excels in reasoning but primarily benefits high-resource languages such as English and Chinese. However, low-resource languages remain underserved due to the dominance of English-centric training data and model optimizations, which limit performance in these languages. This limitation results in unreliable code-switching and diminished effectiveness on tasks in low-resource languages. Meanwhile, local and regional LLM initiatives have attempted to bridge this gap by developing language-specific LLMs that focus on improving local linguistic fidelity. We demonstrate that, with only publicly available datasets and a computational budget of $120, it is possible to enhance the reasoning capabilities of language-specific LLMs to match the level of DeepSeek R1, without compromising their performance on target language tasks.