In-Context Learning

In-context learning is one of the surprising emergent abilities of large language models: by including examples of a task within the prompt itself, the model can pick up on the pattern and apply it to new inputs — even for tasks it wasn’t explicitly trained on. This is different from finetuning, which changes the model permanently; in-context learning only affects behavior for that specific interaction. It’s the mechanism that makes few-shot prompting work, and it allows models to adapt to custom formats, specialized domains, or specific stylistic requirements on the fly. For behavior architects, in-context learning is a powerful technique to keep in mind when deciding whether a training-based solution or a prompt-based solution is the right approach to a behavioral problem.