Temperature and Sampling Quiz

Answers and explanations appear as you take the quiz.

1. What does raising the temperature do to a model's output?
   • Makes it more deterministic and repetitive
   • Permanently changes the model's weights
   • Disables the softmax function
   • Makes it more random and varied, with more chance of unusual tokens
2. Before sampling, the model's raw token scores (logits) are turned into probabilities by:
   • The top-k filter
   • The softmax function
   • Backpropagation
   • The tokenizer
3. What determines the candidate set in top-p sampling, compared with top-k?
   • Top-p always keeps exactly ten tokens
   • Both methods always keep every vocabulary token
   • Top-p uses the smallest leading set reaching a cumulative probability threshold; top-k uses a count
   • Top-k uses a cumulative threshold while top-p uses a fixed count
4. Mechanically, what does temperature actually do to the logits?
   • Divides each logit by T before softmax
   • Adds a random number to each logit
   • Deletes the lowest logits
   • Multiplies the final probabilities by T
5. For a structured-data task, what does choosing more predictable token selection fail to establish?
   • That a decoding setting was chosen
   • That the generated content is correct or valid
   • That token scores can be compared
   • That a model can produce an output
6. How does greedy decoding choose the next token?
   • It literally divides every logit by zero
   • It uniformly samples every vocabulary token
   • It permanently retrains the model
   • It selects a highest-scoring eligible token rather than sampling
7. Two token logits differ by six before temperature scaling. With positive temperature two, what is their scaled difference?
   • Three
   • Twelve
   • Six
   • Zero
8. For probabilities 0.55, 0.30, 0.10, and 0.05, how many leading tokens does top-p 0.80 keep?
   • One
   • Three
   • Two
   • Four