In-Context Learning
The ability of a model to learn patterns from instructions and examples provided inside the current prompt without updating its weights.
In-context learning is the phenomenon where a model adapts to a task based only on the text in the current prompt. The model does not update its internal parameters; it simply uses the examples and instructions in context to infer what to do.
This is one of the most surprising properties of modern LLMs. A single model can translate, summarize, classify, extract data, and write code, all because it can interpret the task from context rather than needing a separate trained model for each job.
How Teams Use In-Context Learning
- Rapid prototyping — test workflows without retraining
- Few-shot examples — teach the desired pattern inside the prompt
- Task switching — reuse one model across many functions
- Light customization — adapt outputs with instructions and context
In-context learning is a big reason general-purpose LLMs are so flexible. It shifts much of the work from model training to prompt design and application architecture.