Temperature
A hyperparameter in generative models that scales the output logits before the softmax function, controlling the randomness and creativity of the output.
Think of It Like This
Like adjusting a dial between a strict, by-the-book bureaucrat (low) and a wild, free-associating poet (high).
A temperature of 0.0 makes the model entirely deterministic, always selecting the highest-probability token. A higher temperature (e.g., 1.0) flattens the probability distribution, making less likely tokens more probable. This is crucial for creative writing tasks, while low temperatures are mandatory for strict tasks like coding or math.