Domain 1 of 6

How a Model Reads a Prompt

The mechanism underneath every technique that follows. Tokens and why position in the window matters, the instructions and examples a model is given, the sampling settings that decide how much the same prompt varies, and the prompt patterns that survive contact with a real workload.

Concepts

~15%

Of the exam

Practice questions

Concepts in this domain

01How a language model reads a promptWhy a model receives tokens and never words, the loop that produces one token at a time from everything before it, and why every token already in the window has a claim on the same attention.02Prompt engineeringThe parts a prompt is made of, the shot based techniques and chain of thought, the patterns underneath agents, the four named risks, and where each syllabus places the material.03Instructions, examples and output formatWhat a worked demonstration teaches that a written rule cannot, how many examples earn their tokens, how to specify the shape of an answer so a program can parse it, and the order the parts of a prompt go in.04Sampling and determinismWhy the same prompt produces a different answer twice, what temperature does to the probability distribution, how top k and top p truncate it before the draw, and why a fixed seed at temperature zero still fails to reproduce a call.05Prompt patterns that holdWhat belongs in a system prompt and what quietly does nothing, the limits of role framing, splitting work inside one call, an honest account of when chain of thought helps, and self consistency with the price of a vote.

Try a question from this domain

A housing association holds 200 support policies, each running to about 1,500 words. An engineer proposes placing all of them in the window of a model that accepts 200,000 tokens, on the grounds that 300,000 words of policy is a small library and the window is large. Working from the published ratio of about four characters of English to a token, what does the arithmetic show?

AAbout 225,000 tokens, since a token is about three quarters of a word and three quarters of 300,000 words is 225,000, so the policies overflow by roughly an eighth.
BAbout 400,000 tokens, since 1,500 words is roughly 2,000 tokens and 200 policies at 2,000 tokens each is 400,000, which is twice the whole window before anything else is counted.
CAbout 300,000 tokens, since English runs to roughly one token a word once punctuation and spacing are counted in, so the policies overflow by half.
DThe policies fit, since the input and the generated answer are counted against separate limits and only the answer has to stay inside the 200,000.

14 questions on this domain.

One per page, with a worked explanation.

Start the set