Fine-Tuning
Fine-tuning adapts a base model to your domain, tone, or task by training on curated examples, beyond what a system prompt alone can reliably enforce.
Designers feel fine-tuning as more consistent voice, format, and terminology, but also as slower update cycles when brand or policy changes.
What it means
Additional training (full or lightweight) on your labeled data so the model’s default behavior skews toward your product’s patterns.
Why designers should care
When prompts and few-shot examples plateau, fine-tuning may be needed, but you must plan versioning, evals, and rollback because behavior shifts in subtle ways.
Example
A healthcare portal fine-tunes on approved patient-facing phrases; the UI still shows “Draft: clinician review required” because tuning reduced but did not eliminate risk.
Common mistakes
- • Expecting fine-tuning to replace guardrails or human review in regulated domains.
- • Shipping a tuned model without A/B UX for regression on edge cases.
- • Confusing fine-tuning with RAG. Retrieval adds facts; tuning changes style and priors.