Definition
In-context learning (ICL) allows large language models to adapt to new tasks without updating their weights. By providing input-output pairs within the prompt context, the model infers the pattern and applies it to new queries. This zero-shot or few-shot capability enables rapid prototyping and flexibility, serving as a powerful alternative to traditional fine-tuning for tasks requiring quick adaptation to novel domains.
Summary
A technique where models learn to perform tasks by observing examples provided in the prompt.
Key Concepts
- Few-Shot Learning
- Zero-Shot
- Prompt Design
- Weight-Free Adaptation
Use Cases
- Quickly testing model capabilities on new datasets
- Dynamic task switching in conversational agents
- Prototyping AI applications without retraining
Code Example
| |