noun
A fine-tuned model is an AI model that has been further trained on a specific dataset after its initial training. This process adapts a general-purpose model (like a large language model) to perform better on niche tasks, such as answering questions in a specific domain or generating more accurate images.
Fine-tuning works by taking a pre-trained model and adjusting its parameters using a smaller, task-specific dataset. This is cheaper and faster than training a model from scratch. For example, a language model fine-tuned on medical literature can answer health-related questions more accurately than a general model.
For users, fine-tuned models offer more precise and relevant outputs, but they may require more computational resources and cost more to deploy. A company using a fine-tuned chatbot for customer support might see a 20% improvement in response accuracy but pay higher hosting fees.
A fine-tuned model helps a travel app suggest personalized itineraries based on user preferences.
In other languages
See also: Transfer Learning · prompt engineering