Fine-tuning
Giving a pre-trained AI model additional training with an organisation's own example data so that it fits a specific task or style.
- 01
Why it matters
Fine-tuning can make a model more accurate in work that needs a consistent format, style or classification. It is usually not the right way to teach frequently changing information, though; retrieval-augmented generation suits that better. Good results depend on clean, representative example data.
- 02
Example
A company fine-tunes a model with two thousand labelled support tickets. The model sorts new tickets into the right category and tickets reach the relevant team faster.
- 03
Common mistake
Trying to teach constantly changing information such as prices and stock through fine-tuning. The model memorises outdated information and needs retraining every time it changes.
- 04
Related terms
Related services and guides
Let's talk about your project.
Tell us what you need; we will define the scope together.