Multimodal Large Language Models for Low-Resource Languages: A Case Study for Basque

Researchers successfully developed multimodal large language models for Basque, a low-resource language, finding that only 20% Basque training data is needed for solid performance. The study demonstrates that specialized Basque language backbones aren't required, potentially enabling MLLM development for other underrepresented languages.

• →Low ratios of Basque multimodal data (around 20%) are sufficient to achieve solid benchmark results.
• →A Basque-specific instructed backbone LLM is not required to build strong MLLMs in Basque.
• →The research creates new training and evaluation image-text datasets for Basque language.
• →Two different LLM backbones were tested: Llama-3.1-Instruct and Basque-adapted Latxa.
• →Resources are being openly released to enable MLLM development for other low-resource languages.