Fine-Tuning
Giving an already-trained AI extra training on specific examples to specialise it.
In everyday terms
A general model can be fine-tuned on thousands of customer-support replies so it writes in your company's style.
For professionals
Continued training on a narrower dataset, full or parameter-efficient (e.g. LoRA). Changes behaviour and style more reliably than it adds facts.
Think of it like…
A qualified doctor doing a specialist residency.
You've already seen it
Custom models, "trained on our data" claims, specialised medical or legal assistants.
Myth vs reality
Myth: Fine-tuning is the best way to teach an AI your company's facts.
Reality: For facts that change, RAG is usually better. Fine-tuning suits style and format.
Quick check
To make an AI always reply in your brand's tone, the best fit is…
- Fine-tuning
- Raising temperature
- A vector database
- Overfitting
Show answer
Fine-tuning: Fine-tuning shapes style and behaviour.
Builds on
Related
Training · Retrieval-Augmented Generation (RAG) · Prompt Engineering