FOUNDATIONS • 17 / 397
Start before LLMs. Understand where AI, ML and neural networks fit.
Test Set
The test set is held-out data used for a more final, unbiased check of performance.
Think of it like
Think of Test Set as one layer in a very large toolbox: you only need to understand what job that layer performs.
Real life
A phone spam filter, recommendation engine, fraud detector or voice assistant may use Test Set somewhere inside.
SRE lens
A model should not have trained on the incidents used for its final evaluation.
Remember thisThe test set is held-out data used for a more final, unbiased check of performance.