Language Models are Few-Shot Learners

https://arxiv.org/abs/2005.14165