Adjust the temperature to see how it affects the probabilities of different tokens in the model's output.