Overview
Few-shot learning aims to mimic the human ability to learn new concepts from very little data. It is a middle ground between one-shot and traditional many-shot learning.
Meta-Learning
FSL is often approached through 'meta-learning' or 'learning to learn,' where the model is trained on a variety of tasks so it can quickly adapt to new ones.
Use in LLMs
Providing a few examples in a prompt (few-shot prompting) is a common way to guide LLMs to produce the desired output format or style.