OpenSeal: Good, Fast, and Cheap Construction of an Open-Source Southeast Asian LLM via Parallel Data
This provides a transparent, open-source solution for improving multilingual LLM performance in Southeast Asian languages, though it is incremental as it builds on existing parallel data methods.
The researchers tackled the problem of poor performance of English-centric large language models on low-resource Southeast Asian languages by using parallel data for continual pretraining, resulting in OpenSeal, an open-source model built with 34.7B tokens and 180 GPU hours that rivals existing models of similar size.
Large language models (LLMs) have proven to be effective tools for a wide range of natural language processing (NLP) applications. Although many LLMs are multilingual, most remain English-centric and perform poorly on low-resource languages. Recently, several Southeast Asia-focused LLMs have been developed, but none are truly open source, as they do not publicly disclose their training data. Truly open-source models are important for transparency and for enabling a deeper and more precise understanding of LLM internals and development, including biases, generalization, and multilinguality. Motivated by recent advances demonstrating the effectiveness of parallel data in improving multilingual performance, we conduct controlled and comprehensive experiments to study the effectiveness of parallel data in continual pretraining of LLMs. Our findings show that using only parallel data is the most effective way to extend an LLM to new languages. Using just 34.7B tokens of parallel data and 180 hours on 8x NVIDIA H200 GPUs, we built OpenSeal, the first truly open Southeast Asian LLM that rivals the performance of existing models of similar size.