Synthetic data: definition

Synthetic data is artificially generated examples used to train or test a model when real data is scarce, sensitive, or imbalanced.

Sometimes you do not have enough real examples, or the real data is too sensitive to use. Synthetic data fills the gap: realistic generated examples that carry the patterns you need without the privacy or scarcity problems.

It is useful for rare cases (fraud, edge conditions) and for protecting personal data, but it has to be validated against reality, because a model trained on unrealistic synthetic data learns the wrong lessons.

Related terms

Fine-tuning: Fine-tuning is further training a base model on your own examples so it learns a specific style, format, or task more reliably than prompting alone can achieve.
Evaluation (evals): An evaluation, or eval, is a repeatable test that measures how well an AI system does its job on real examples, so you can tell whether a change made it better or worse.
Hallucination: A hallucination is when a model states something false or invented as if it were fact, because it predicts plausible-sounding text rather than looking anything up.