Claim record · edition 1.0.0
si-002Established result
Sufficiently large autoregressive language models perform tasks from instructions and examples supplied in context, without gradient updates.
In-context few-shot behaviour was the central reported finding of the GPT-3 paper and is reproduced widely.
Limits of this claim
Reported on the task distributions studied. It is not a claim that in-context learning generalizes to arbitrary tasks, and it says nothing about reliability.
Supporting source records · 1
primary paperContent verified
Language Models are Few-Shot Learners
Brown et al. · 2020
Introduced in-context few-shot learning as an emergent property of scale.
Source record →