How to use these explanations

Read the chapters in order; they are arranged from the most traditional style of testing to the most modern. Bold linked terms like ground truth jump to the glossary, where every key idea is explained from scratch. Foundational terms that belong to the earlier folders (such as what a benchmark is at heart, or what an Elo rating is) are linked across to those folders rather than repeated here.

You do not need anything from the earlier folders memorized. The one idea worth carrying in is that a language model is a next-word predictor that has been trained and tuned; everything here is about how we grade the result.