What is the name for the map of connections between words, built by tallying word usage?	The Transition Matrix (or Frequency Table)
What process is used to select the next word based on the current context in a language model?	Sampling
What is the result of normalizing each row of a transition matrix?	A probability distribution of what word is likely to come next
In a Markov chain model, how is the next word predicted?	By looking up a row in the transition matrix (a simple table lookup)
What complex mechanism do modern Large Language Models (LLMs) use instead of a simple table lookup?	A forward pass through billions of parameters (using attention, embeddings, etc.)
What does Top-k sampling restrict when generating text?	Only the N most likely words survive (e.g., only the top 3).
How does Top-p sampling (or nucleus sampling) restrict word choices?	It keeps the smallest set of words whose cumulative probabilities sum to a specified threshold (e.g., 90%).
What parameter controls the randomness of the sampling process in an LLM?	Temperature (A higher temperature increases randomness).
What historical model, used in early phone prediction, relied on counting sequences of words?	N-gram models
What is the key difference in context reading between a Markov chain and an LLM?	Markov chains read only the last word; LLMs read the entire conversation so far.