Definition
Fine-tuning is the process of adapting a previously trained base model to perform better for a particular task, domain, or response style.
A base language model may have broad language capabilities. It can then be trained on a more specific set of examples, such as question-and-answer pairs. For example, a model could be adapted to:
- Answer customer-service questions
- Work with medical terminology
- Handle financial-domain content
- Follow a particular response style
During fine-tuning, the model’s internal weights are adjusted so that it produces responses that better match the desired behavior.
Why It Matters
Fine-tuning and distillation help you evaluate model specialization strategies — when it’s worth adapting a model versus relying on prompting alone.