noun
AI fine-tuning is the process of taking a pre-trained AI model and adapting it to perform a specialized task more accurately. Instead of training a model from scratch, fine-tuning leverages an existing model and refines it with additional data relevant to the new task.
This process typically involves feeding the model new examples, adjusting its parameters, and sometimes freezing certain layers to prevent overfitting. Fine-tuning is common in language models, image generators, and other AI tools where general models need to be tailored for specific applications.
For users, fine-tuning can mean better results but also higher costs and technical complexity. For example, a company fine-tuning a language model for customer service might spend $500–$2,000 to improve response accuracy by 10–15%. However, it also requires expertise in AI training and access to quality data.
A startup fine-tunes a text-to-image model to generate better product mockups for e-commerce.
In other languages
See also: Transfer Learning