JASMINE: Arabic GPT Models for Few-Shot Learning
This addresses the gap in understanding autoregressive models for Arabic, a language with over 400 million speakers, by providing models and a benchmark for evaluation, though it is incremental as it applies existing methods to a new linguistic context.
The authors tackled the lack of generative pretraining models for Arabic by introducing JASMINE, a suite of Arabic GPT models ranging from 300 million to 6.7 billion parameters, pretrained on 235 GB of text, and showed powerful performance in few-shot learning on various NLP tasks.
Scholarship on generative pretraining (GPT) remains acutely Anglocentric, leaving serious gaps in our understanding of the whole class of autoregressive models. For example, we have little knowledge about the potential of these models and their societal impacts in diverse linguistic and cultural settings. We alleviate this issue for Arabic, a wide collection of languages and dialectal varieties with more than 400 million population, by introducing JASMINE. JASMINE is a suite of powerful Arabic autoregressive Transformer language models ranging in size between 300 million-6.7 billion parameters pretrained on a large and diverse dataset (~ 235 GB of text). We also carefully design and release a comprehensive benchmark for both automated and human evaluation of Arabic autoregressive models, with coverage of potential social biases, harms, and toxicity. Using our novel benchmark, we evaluate JASMINE extensively showing powerful performance intrinsically as well as in few-shot learning on a wide range of NLP tasks. We aim to responsibly release our models and evaluation benchmark with interested researchers, along with code for experimenting with them.