Language Models are Few-Shot Learners

The coauthored GPT-3 paper examines in-context learning, evaluation and model limitations. Mann contributed to model and training infrastructure, data preparation and downstream evaluation.