TituLLMs: A Family of Bangla LLMs with Comprehensive Benchmarking

TL;DR

This paper introduces TituLLMs, the first large pretrained Bangla language models, along with comprehensive benchmarking datasets, demonstrating their performance and laying the foundation for low-resource language adaptation.

Contribution

We developed TituLLMs, the first Bangla LLMs, and created benchmarking datasets to evaluate their performance, facilitating research in low-resource language modeling.

Findings

TituLLMs outperform initial multilingual models in Bangla tasks.

Benchmarking datasets reveal language adaptation complexities.

Models and datasets are publicly available for research.

Abstract

In this paper, we present TituLLMs, the first large pretrained Bangla LLMs, available in 1b and 3b parameter sizes. Due to computational constraints during both training and inference, we focused on smaller models. To train TituLLMs, we collected a pretraining dataset of approximately ~37 billion tokens. We extended the Llama-3.2 tokenizer to incorporate language- and culture-specific knowledge, which also enables faster training and inference. There was a lack of benchmarking datasets to benchmark LLMs for Bangla. To address this gap, we developed five benchmarking datasets. We benchmarked various LLMs, including TituLLMs, and demonstrated that TituLLMs outperforms its initial multilingual versions. However, this is not always the case, highlighting the complexities of language adaptation. Our work lays the groundwork for adapting existing multilingual open models to other low-resource languages. To facilitate broader adoption and further research, we have made the TituLLMs models and benchmarking datasets publicly available (https://huggingface.co/collections/hishab/titulm-llama-family-6718d31fc1b83529276f490a).