In-Context Learning

In-context learning (ICL) is a capability of large language models whereby the model adapts its behavior or learns a new task purely from examples or instructions provided within the prompt — without any updates to the model's underlying weights. Rather than retraining, the model uses patterns in the provided context (demonstrations, instructions, or examples) to infer the task and produce appropriate outputs.

In commerce and enterprise applications, ICL enables teams to quickly prototype AI behaviors — such as classifying product returns, extracting structured data from invoices, or drafting customer responses — by supplying a handful of labeled examples in the prompt. This dramatically reduces the time-to-value compared to traditional fine-tuning, though it trades off some consistency and reliability for speed and flexibility. Effective ICL design requires careful prompt construction and awareness of context window limits.