(B) To evaluate the model's accuracy.
The testing dataset is used to evaluate the performance of a trained model by comparing its predictions against known expected values.
Source: Chapter 3, Section 3.2 (Train-test split)
The key idea from the passage is that the test dataset's input is fed to the trained model, predictions are made, and those predictions are compared to expected values — this is evaluation. The training dataset is what teaches/trains the model (ruling out option A). Options C and D are unrelated to the purpose of a test split.