TL;DR
A prompting paradigm where a large language model learns to perform a task by observing examples provided directly within the prompt, without modifying any underlying weights.
This capability allows pre-trained models to adapt to new tasks instantly at inference time. Instead of executing resource-intensive fine-tuning, the user supplies a few input-output pairs or natural language instructions inside the prompt context. The model leverages its pre-existing attention patterns to recognize the task's structure and generalize to the new target query.
Why this matters for your business
It enables rapid prototyping and highly customizable user experiences by allowing non-technical users to program models through examples alone.
