LanguageModelsareFew-ShotLearners

Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous nonsparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks andfew-shot demonstrations specified purelyviatextinteraction withthemodel.