See, humans respond very differently when that happens. The failure to do what humans do when they don’t understand something or know something is frequently what fails LLMs at the TT.