analytics-dictionary / term / ai
Fine-tuning
Fine-tuning is training an existing model further on your own examples, to specialise its behaviour or style.
Written by Reflective Data experts · updated Oct 5, 2026
It can improve consistency on a narrow task, but it is often unnecessary: good prompts, retrieval and tool use solve many problems at lower cost and with easier updates.
Try prompting and retrieval first, and fine-tune only if evaluations show a gap.
related
AI evaluations (evals)Large language model (LLM)Prompt engineering
// how we can help
We build company brains, agent harnesses and AI bots on top of data you can trust. AI services → · The data platform behind it