in-context

Definition

Behaviour changed by what is in the prompt rather than by any weight update. In-context learning — adapting from a handful of examples in the input — is what made large language models usable without per-task training, and its cost profile mirrors fine-tuning's: nothing to train, but every example is paid for on every call and bounded by the context window. Whatever is learned lasts only for that forward pass.

Explainers using this term

• Paper explained: BDH-CQ — an AI that thinks without words. Recurrent memory plus latent reasoning resets ARC's cost frontierBDH-CQ: In-Context Learning with Recurrent Latent Reasoning