Fine-Tuning
The process of further training a pre-trained AI model on a smaller, domain-specific dataset to adapt it to a particular task or style.
Also known as: model fine-tuning, domain adaptation, supervised fine-tuning
Fine-tuning is the process of taking a pre-trained AI model and continuing to train it on a smaller, task-specific dataset. The pre-trained model already knows language and general patterns from its large-scale training. Fine-tuning teaches it to apply that knowledge to a specific domain, task format, or response style.
For example, a general language model fine-tuned on medical Q&A data becomes more accurate at medical questions. A model fine-tuned on code review examples becomes more useful for code review tasks. A model fine-tuned on customer service transcripts learns the tone and format expected in customer service interactions.
Fine-tuning datasets need to be much smaller than pre-training datasets but much higher quality. A few thousand well-constructed examples often outperform tens of thousands of poorly structured ones. The format is typically JSONL: one JSON object per line, each with an input prompt and an expected output.
For example, a general language model fine-tuned on medical Q&A data becomes more accurate at medical questions. A model fine-tuned on code review examples becomes more useful for code review tasks. A model fine-tuned on customer service transcripts learns the tone and format expected in customer service interactions.
Fine-tuning datasets need to be much smaller than pre-training datasets but much higher quality. A few thousand well-constructed examples often outperform tens of thousands of poorly structured ones. The format is typically JSONL: one JSON object per line, each with an input prompt and an expected output.