AI Glossary
Fine-tuning
model fine-tuning, fine-tuning
Fine-tuning is the further training of a ready-made model on your own set of examples, so it handles a specific task or style better. It changes the model's weights, unlike prompting.
- It changes the model's weights based on your own set of examples.
- It makes sense when prompting and few-shot fall short and the task is repeatable.
- It requires prepared data, the cost of training, and a fresh round of evaluation.
Fine-tuning further trains a ready-made model on your own set of examples, adjusting the model's parameters to a specific task — without training from scratch. You reach for it when, despite a polished prompt, the model fails to hold consistently to a required style, format, or knowledge of a narrow domain.
In practice you turn to it only once prompt engineering and few-shot have hit their limit. Fine-tuning requires gathering and cleaning training data, bearing the cost of training, and re-assessing quality, because changing the weights can improve some cases while degrading others.
Related terms
In guides