Positive temperature scaling will not turn the second-ranked candidate token into the first, nor will it judge factual truth or falsehood; lowering the temperature to 0 only causes wrong answers to appear repeatedly with greater certainty.