AI & Machine Learning Temperature, Top-k & Top-p Quiz

Sampling settings decide how predictable or creative a model's output is. This quiz checks what temperature does to the probabilities, how top-k and top-p (nucleus) sampling filter tokens, and which settings suit which tasks, with a few small numbers to work out.

Medium 10 questions ~6 minutes Pass at 80%

Choose a mode

Playing as a guest: sign in to keep your XP and streak.

Sample question

Question 1 of 10

What does temperature do mathematically?

Show the answer

B. It divides the logits by the temperature before softmax, sharpening or flattening the probabilities Probabilities are softmax(logits / T). A temperature below 1 makes the distribution peakier, above 1 makes it flatter; the weights and the input stay the same.

What this quiz covers

  • Temperature as dividing the logits before softmax
  • Low vs high temperature, and temperature near 0
  • Top-k and top-p (nucleus) sampling with renormalization
  • Choosing settings for extraction vs brainstorming
  • Repetition and frequency penalties

Questions people ask

Do I need an account?

No. Anyone can play. Sign in to keep your XP, streak, badges and history.

Can I retake it?

Yes. Your best score is kept, and the questions and answers are shuffled each time.