CLJan 24, 2024

Large Malaysian Language Model Based on Mistral for Enhanced Local Language Understanding

arXiv:2401.13565v35 citationsHas Code
Originality Synthesis-oriented
AI Analysis

This work addresses the need for improved language models for Malaysian language users, though it is incremental as it builds on existing Mistral 7B with new data and tuning.

The authors tackled the problem of enhancing local language understanding by pretraining Mistral 7B on a Malaysian dataset, resulting in models with extended context lengths that showed superior performance on a Malay grammar test set compared to ChatGPT3.5 and Claude 2.

In this paper, we present significant advancements in the pretraining of Mistral 7B, a large-scale language model, using a dataset of 32.6 GB, equivalent to 1.1 billion tokens. We explore the impact of extending the context length, releasing models with context lengths of 4096 and 32768 tokens, and further refining performance with a specialized 16384 context length instruction-tuned model, we called it Malaysian Mistral. Our experiments demonstrate the efficacy of continue pretraining and the influence of extended context lengths on Mistral 7B's language understanding capabilities. Additionally, we release a model specifically tuned with a 16384 context length instruction, showcasing its potential for capturing nuanced language intricacies. Furthermore, our research contributes to the benchmarking of Malaysian Mistral against prominent language models, including ChatGPT3.5 and Claude 2. We present compelling results indicating Malaysian Mistral's superior performance on Tatabahasa (Malay grammar) test set, particularly when fine-tuned with instructions. All models released at https://huggingface.co/collections/mesolitica/malaysian-mistral-7b-6528f2ec825f4bba46c1700c

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes