Which term is defined as 'a parameter that influences randomness in generated outputs'?

Which term is defined as 'a parameter that influences randomness in generated outputs'?

• A. Specificity
• B. Temperature
• C. False Positive
• D. True Negative

Answer: (A)

Temperature is the parameter that controls randomness in generated outputs. It shapes how the model samples from its predicted distribution: lower temperatures produce more deterministic, focused results by favoring the most probable tokens, while higher temperatures increase randomness and creativity by allowing less probable tokens to be chosen more often. Specificity, false positives, and true negatives are all evaluation metrics for binary classification, not controls on generation or randomness. So the term that best fits the description is temperature.

Temperature is the parameter that controls randomness in generated outputs. It shapes how the model samples from its predicted distribution: lower temperatures produce more deterministic, focused results by favoring the most probable tokens, while higher temperatures increase randomness and creativity by allowing less probable tokens to be chosen more often. Specificity, false positives, and true negatives are all evaluation metrics for binary classification, not controls on generation or randomness. So the term that best fits the description is temperature.