This new V* (guided visual search) model & paper is actually a big deal in my opinion GPT-4V could *not* solve this google recaptcha I had been testing. But now... with the help of the guided V* model it could find the final traffic light