Definition
An evaluation dataset is a set of examples kept out of training and used only to measure how well a model performs. Because it must stay unseen, buyers value fresh, private evaluation data that no model has trained on.
Why it matters when licensing company data
Public benchmarks leak into training data over time, which makes their scores less trustworthy. Real company records, such as resolved support cases with known correct outcomes, can become private evaluation sets for a specific task, and are often licensed under stricter terms than training data.
Source
Google Machine Learning Glossary: test set. Definition reviewed November 7, 2026 by Origin Data Partners.