522 questions comprise the new Humanity’s Sixth Sense benchmark created by Elorian and Scale AI. The test measures how models interpret implied information in visual scenes, such as whether a bookcase can fit more books. Current models struggle with these intuitive tasks.