Few-shot:

A technique where a model can learn to perform a new task or make predictions with limited training examples or labeled data.