Home/
Part XIII — Expert Mode: Systems, Agents, and Automation/42. Fine-Tuning vs Prompting vs Retrieval (Decision Framework)/42.1 When fine-tuning helps (style, format, narrow tasks)
42.1 When fine-tuning helps (style, format, narrow tasks)
Overview and links for this section of the guide.
On this page
Style Transfer
Fine-tuning is great for "vibes." - "Write like a 19th-century poet." - "Write customer support emails in our brand voice (polite but firm)." - "Write Python code that follows our internal style guide (no classes, only functions)."
Prompting this takes 1000 tokens of examples. Fine-tuning burns it into the weights.
Format Enforcement
If you need the model to output a custom binary format or a very obscure XML schema that it messes up with prompting, fine-tuning on 500 examples usually fixes it.