Definition
What is SFT data
SFT data is supervised fine-tuning data: prompt and target-response pairs used to teach a model a behavior, format, or domain style. Expert SFT is authored or heavily edited by practitioners.
What is sft data?
SFT data is supervised fine-tuning data: prompt and target-response pairs used to teach a model a behavior, format, or domain style. Expert SFT is authored or heavily edited by practitioners.
Why it matters
The model will imitate the target distribution, including its fluent mistakes, if those mistakes are in the pairs.
Example
A pharmacist writes reference answers for drug-information questions, including what the model should refuse to invent.
When it is used
You need the model to produce a specific professional form of answer, not only to pick the better of two drafts.
Common mistakes
- Putting benchmark items into SFT and later reporting the same items as independent scores
- Experts writing without a rubric
- No ownership of dataset versioning
What kind of experts do you need?
Share the profession, specialty, experience, location, headcount, hours, duration, and project description. We assemble the expert capacity.
- Need 25 licensed nurses for a clinical AI evaluation
- Need 15 attorneys in a specific practice area
- Need 20 PhD scientists for benchmark creation
- Need 30 senior software engineers for code evaluation