Language models · 2020 · Tom B. Brown et al.

Language Models are Few-Shot Learners

Show that a sufficiently large autoregressive model can perform many tasks from instructions and examples in context without updating its weights.

The central move

Show that a sufficiently large autoregressive model can perform many tasks from instructions and examples in context without updating its weights.

Why it had to exist

NLP systems were commonly fine-tuned separately for each benchmark. Scaling raised the possibility that one pretrained model could infer task behavior from the prompt itself.

Where it leads

Autoregressive scale → in-context learning → prompting and general-purpose language-model interfaces.

Study the guided reading in Bits →