What is "in-context learning" in the realm of Large Language Models (LLMs)? Teaching a model through zero-shot learning Modifying the behavior of a pretrained LLM permanently Training a model on a diverse range of tasks Providing a few examples of a target task via the input prompt