Generative AI Leader Practice Question: The primary purpose of the temperature parameter…

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  It sets the maximum number of tokens in the response

  Why it's wrong here

  That limit is set by max_tokens (or max output tokens), a hard generation cap. Temperature instead rescales logits before softmax, altering how randomly the next token is sampled. Confusing the two is easy since both shape the response, but only one bounds its length.

• ✓

  It controls the randomness or creativity of the output

  Why this is correct

  Temperature scales the probability distribution over the model's next-token predictions before sampling. A low value sharpens the distribution, favouring high-probability tokens for deterministic, factual output; a high value flattens it, giving unlikely tokens more chance and producing varied, creative text. This directly governs the randomness the question asks about.

• ✗

  It determines the number of most likely tokens considered at each step

  Why it's wrong here

  Top-k sampling, not temperature, caps how many candidate tokens are retained per step. Temperature rescales the logits before softmax, flattening or sharpening the probability distribution to control randomness. It is tempting because both parameters tune output diversity, but temperature never restricts the candidate set itself.

• ✗

  It sets the cumulative probability threshold for token selection

  Why it's wrong here

  Temperature rescales the logits before the softmax, altering the shape of the probability distribution; it does not set a cumulative probability cutoff. That cutoff describes top-p (nucleus) sampling, which truncates the token set, making this option tempting when the two sampling controls are confused.