Low-resource languages are difficult to train from scratch due to limited training data. In the attached poster, we show that using minimal data for pre-training and instruction tuning, we are able to achieve better accuracy than other open source #LLMs in the respective

→ View original post on X — @sambanovaai,