What is Fine-Tuning?
Fine-tuning is the process of taking a pre-trained AI model and continuing its training on a smaller, task-specific dataset to adapt it for a particular use case.
It starts with a model already trained on a large general dataset, then updates the model's parameters using new labeled data relevant to the target task.
Training typically uses a lower learning rate to make small adjustments rather than overwriting the original knowledge, helping the model retain useful features while specializing.
This approach is a core part of transfer learning and is commonly applied in NLP and computer vision when labeled data for the new task is limited.
Example
A language model pre-trained on internet text can be fine-tuned on customer support chat logs so it learns to answer questions in a company's specific tone and domain.
Why it matters
Fine-tuning lets organizations adapt powerful foundation models to specialized needs without the enormous cost of training from scratch, making advanced AI practical for many applications.
Frequently asked questions
Training from scratch builds a model using only the target dataset and random initial weights, while fine-tuning starts from weights already learned on a much larger dataset.
Related terms
Transfer learning is a machine learning method that reuses a model trained on one task as the starting point for a different but related task.
Overfitting happens when a machine learning model learns the training data too closely, including its noise and quirks, so it fails to perform well on new, unseen data.
The learning rate is a hyperparameter that controls the size of the steps an optimization algorithm takes when updating a model's parameters during training.
An epoch is one complete pass of a machine learning model through the entire training dataset during training.
Supervised learning is a machine learning method where a model is trained on data that already has correct answers attached, so it can learn to predict those answers for new data.
Batch size is the number of training examples processed together in a single forward and backward pass during model training.