LLM & GenAI Fundamentals: the practice test
Tokens, context windows, prompting vs RAG vs fine-tuning, hallucinations, evals, cost and latency: the conceptual core every AI-native FDE loop probes first. This test drills exactly that: 12 easy, 10 medium and 15 hard questions, every one explained, every explanation linking into the worked material.
Sample questions, answered
Models read and write subword units, not words: 'unbelievable' may be three tokens while 'the' is one. The practical rule of thumb is roughly 4 characters or 0.75 words per token in English, and it is worth memorizing because tokens drive all three numbers a customer cares about: cost, latency, and how much fits in the context window. Per-word pricing is the trap answer; token counts diverge sharply from word counts on code and non-English text.
A language model is trained to continue text plausibly. Fluency and truth are correlated in the training data, so it is often right, but nothing in next-token prediction verifies claims against a source of truth, so when the model does not know, it produces something shaped like an answer. The other explanations name real but secondary factors: bad data, sampling and cutoffs change how often it happens, not why it is possible. This framing also tells you why grounding plus permission to say 'I don't know' is the fix that works.
Go deeper than the quiz
A practice test measures recall. The material it draws from teaches the reasoning: