Fine-tuning is additional training that adapts a pre-trained model for a more specific behavior or task.
Fine-tuning is additional training that adapts a pre-trained model for a more specific behavior or task.
It starts from an existing model and continues training on focused examples. Some methods update model parameters directly; others train adapters.
Think of it like specialist training after general education. The learner builds on broad skills instead of starting from zero.
A base model is exposed to target examples. Training updates selected values, then evaluation checks the result.
Input: A customer ticket
System: Examples pair tickets with desired labels
Output: More consistent categories
Input: A draft request
System: Examples demonstrate tone and format
Output: A more consistent draft
Input: A document
System: Examples show the target schema
Output: More predictable fields
Fine-tuning: Targeted additional training
Pre-training: Broad initial training
Fine-tuning: Changes learned behavior through training
Prompting: Changes instructions at runtime
Fine-tuning: Adapts model behavior
RAG: Retrieves external information at query time
Explore next
Fine-tuning is additional training that makes a pre-trained model more consistent for a specific behavior or task.