Test Set
Domain: Data & Datasets
🎯 Motivation
Data the model has never seen, used only for final evaluation. Provides an honest performance estimate.
📋 Concrete Example
Test set: 2,000 new examples. Model scores 85%. This is the honest estimate of real-world performance. The model never trained on or tuned with this data.
🔗 Analogy
The real HKDSE exam - questions the student has never seen. It provides an honest measure of their ability. The test set is the real exam for the model.