Examples
Worked examples
- Is an instance
LoRA fine-tuning of Llama 3 on a corpus of cardiology guidelines to build a domain assistant
Counter-examples
Looks similar, but isn't
- Not an instance
Few-shot prompting (showing examples in the prompt) is not fine-tuning — model weights are unchanged
Editorial commentary
Variants include full fine-tuning, parameter-efficient methods (LoRA, adapters), instruction tuning, and RLHF. For scholarly disclosure, the base model, fine-tuning data, and method should be reported; fine-tuning materially changes what the model can be expected to do.
References
- Howard and Ruder 2018 ‘Universal Language Model Fine-tuning’ ACL
- Hu et al. 2022 ‘LoRA: Low-Rank Adaptation’ ICLR
Also known as
Model adaptation · Domain adaptation (LLM)
Machine-readable encodings
Use in your systems
xml
<role vocab="credit"
vocab-identifier="https://casrai.org/dictionary/"
vocab-term="Fine-tuning"
vocab-term-identifier="https://casrai.org/dictionary/term/fine-tuning" />json
{
"@context": "https://schema.org",
"@type": "DefinedTerm",
"name": "Fine-tuning",
"identifier": "https://casrai.org/dictionary/term/fine-tuning",
"description": "The process of further training a pre-trained foundation model on a smaller, task-specific or domain-specific dataset, updating some or all parameters, to specialise its behaviour while retaining general capability.",
"inDefinedTermSet": "https://casrai.org/dictionary/domain/generative-ai-use-and-disclosure/",
"url": "https://casrai.org/dictionary/term/fine-tuning",
"sameAs": [
"Model adaptation",
"Domain adaptation (LLM)"
],
"license": "https://creativecommons.org/licenses/by/4.0/"
}
