In-Context Learning
A model performing a task from examples in the prompt, with no weight update.
Overview
In-context learning is a model's ability to infer a task from examples supplied in the prompt and apply it to a new input, without any parameter change. It solves the problem that training a model per task is expensive when most tasks can be demonstrated in a paragraph. It exists because it emerged from scale rather than being designed: the capability appeared in large models that were never trained for it.
Technology Graph
How In-Context Learning connects to the rest of the stack.
Tools
Software that implements or integrates with In-Context Learning.