What are the three main stages of how Claude generates text?	Tokenization, the prediction phase, and the sampling phase.
What happens during Claude's prediction phase?	It decides what possible next words could come next and assigns a probability to each option.
What is the function of the sampling phase in text generation?	A token is actually chosen based upon the probabilities assigned during the prediction phase.
What is the 'temperature' parameter, and what range is it typically in?	A decimal value between zero and one that influences the exact distribution of probabilities.
What kind of output is produced when the temperature is set to zero?	Deterministic output, where the model always selects the token with the highest initial probability.
How does increasing the temperature affect token selection?	It increases the chances of selecting a token that has a lower initial probability.
For what types of tasks should a low temperature be used?	Tasks requiring little randomness or creativity, such as data extraction.
For what types of tasks should a high temperature be used?	Creative tasks such as brainstorming, writing, or generating jokes.
Does increasing the temperature guarantee a dramatically different output?	No, it only increases the *chance* of getting a different token.