Knowledge Maxxing← Back to topic library
AI MaxxingIntermediate75 sec
Fine-tuning
Fine-tuning is additional training performed on a pretrained model using a smaller, more specific dataset to adjust its behavior for a particular use case.
Prerequisite: Pretraining
What matters
Because fine-tuning starts from a model that already has broad language ability, it needs far less data than training from scratch to specialize the model's behavior.
Example
A general language model fine-tuned on customer support transcripts learns to answer in that company's tone and cover its specific products.
Common misconception
Fine-tuning does not erase everything the model learned during pretraining; it adjusts the model's existing knowledge rather than replacing it.
Recall question
Why does fine-tuning typically need much less data than pretraining?
Keywords and related concepts
fine-tuningspecializationtraining datapretraining