Repetition Penalty and Decoding Controls Quiz

Answers and explanations appear as you take the quiz.

1. What does raising the 'temperature' parameter do to a model's output?
   • Makes it more random and diverse
   • Makes it more deterministic and focused
   • Permanently changes the model's weights
   • Speeds up generation
2. How does top-p (nucleus) sampling decide which tokens to consider?
   • It keeps a fixed number of tokens regardless of probability
   • It keeps the smallest set of tokens whose probabilities sum to a threshold
   • It keeps only the single most likely token
   • It keeps tokens with probability above 50%
3. At what stage does a repetition penalty typically act?
   • On the raw text after generation
   • On the training data
   • On the logits before the softmax
   • On the tokenizer vocabulary
4. In OpenAI’s text-generation API, how do presence and frequency penalties differ?
   • Presence penalty scales with token count; frequency penalty is flat
   • They are identical
   • Both only apply to the first token
   • Frequency penalty scales with how often a token appeared; presence penalty applies once if it appeared at all
5. For code completion where deterministic sampling is desired, which setting best matches the guide?
   • Temperature near zero
   • Very high temperature
   • High repetition penalty above 2.0
   • Top-p of 1.0 with high temperature
6. What does top-k sampling do?
   • Keeps tokens summing to probability k
   • Keeps only the k most probable tokens as candidates
   • Multiplies all logits by k
   • Removes the k least likely tokens from training
7. With a repetition penalty above 1, how does a standard logits processor adjust a previously seen token?
   • It sets every repeated-token logit to zero
   • It adds the penalty to every token equally
   • It divides a positive logit by the penalty and multiplies a negative logit by it
   • It removes the token from the vocabulary permanently
8. How does top-k differ from top-p sampling?
   • Top-k uses cumulative probability while top-p fixes a count
   • Top-k changes model weights while top-p changes the tokenizer
   • They are both deterministic decoding methods
   • Top-k keeps a fixed number of candidates; top-p keeps a probability-mass-based set