Asked about Running a fine-tune properly
Where do the examples come from?
An answer that works
Rarely from writing them by hand, which stalls at a few hundred. The two that work are a larger model's output filtered through your own evaluation, which is cheap and bootstraps the first version, and corrected production traffic, which is the version you keep. Each brings a defect worth naming: the distilled data inherits the teacher's mistakes, so the filter is not optional, and log-derived data covers only what the current system already handles well, so the failures you most need have to be sampled deliberately.
Reading an answer is not being able to give one. The section that teaches this is Running a fine-tune properly, 6 hours of Fine-tuning.
Learn this properlyAlso asked about Fine-tuning
When would you fine-tune rather than use RAG?What is LoRA?What is the cost of fine-tuning beyond the training run?What is distillation?How much data do you need to fine-tune?What is catastrophic forgetting?How do you know the fine-tune actually beat the prompt?You have five fine-tunes of one base model to serve. Merged or separate adapters?
All 13 questions on Fine-tuning