Deep Learning · Level 2 of 5
Fine-tuning
Continuing training of a pretrained model for a target task or dataset.
Updating parameters adapts the model; it can also damage previous capabilities.
Example
A pretrained language model is trained further on support examples.
Listen to the definition and example
Audio transcript
Fine-tuning. Continuing training of a pretrained model for a target task or dataset. Updating parameters adapts the model; it can also damage previous capabilities. For example: A pretrained language model is trained further on support examples.
Explore this concept
A useful analogy
Adjusting an already trained skill through further practice.
Why it matters
This helps explain why a neural model learns well, becomes unstable, or fails to generalize.
Technical detail
Training continues from pretrained θ on a new objective or dataset; some methods update only a small subset or additional parameters.
Common misconception
Putting documents in a prompt does not update weights and is not fine-tuning.
Start with
Related concepts
Quick recall question
Try answering before looking back at the definition.