How do you decide which synthetic examples belong in a fine-tuning dataset?
A generated answer can be correct and still add little training value. Audit the target, the available evidence and the decision it teaches before training.
synthetic data
1 post with the tag “synthetic data”