Few Shot Learning
In plain terms
Teaching a model a task with just a few examples.
Going deeper
Few-shot learning gives a model a handful of examples in the prompt or training so it generalizes to the task without large datasets. It learns from minimal demonstration.
Under the hood
In-context few-shot prompting steers LLMs at inference; contrasts with zero-shot and fine-tuning.
Examples
In-context prompting, LLM examples, low-data tasks, classification.
Think of it like...
Like showing someone two or three samples and they grasp the pattern.