Displays the raw, unnormalized scores (logits) the model assigns to every possible next token in the vocabulary before probabilities are calculated.