Hmm by the abstract and the question list they didn't measure terse fluff-less prompts?