What This Research Area Covers
Fine-tuning research studies methods for adapting a pretrained model to a specific task or domain through additional, targeted training, including techniques for doing this efficiently without needing to retrain the entire model from scratch.
Why It Matters
Fine-tuning lets an organization specialize a general-purpose model for its specific use case, tone, or domain, and efficient fine-tuning methods make this practical without the enormous cost of full retraining.
Current Research Directions
Parameter-efficient fine-tuning methods (adjusting only a small subset of a model's parameters), techniques for avoiding "catastrophic forgetting" of general capability during specialization, and combining fine-tuning with other techniques like RAG are active areas.
Related Pages
Frequently Asked
Is fine-tuning the same as prompt engineering?
No — prompt engineering shapes behavior through instructions at query time; fine-tuning changes the model's underlying weights through additional training.
What is parameter-efficient fine-tuning?
A family of techniques that adjust only a small subset of a model's parameters rather than the entire model, reducing cost significantly.
Should I fine-tune or use RAG for my use case?
See our Fine-Tuning vs RAG page for a practical decision framework.
Where can I learn the basic concept?
See our Fine-Tuning vs RAG page.