← Back to Glossary
Instruction Tuning
Fine-tuning a base model on prompt-and-response pairs so it follows user commands instead of just completing text.
Why it exists
Base pretrained LLMs tend to complete text (e.g. responding to a question with more questions). Instruction tuning turns them into conversational assistants.
How it works
Trains the model on supervised dataset pairs containing explicit instructions and ideal completions (Supervised Fine-Tuning - SFT).
Related terms
Where you'll use this
roadmap