We will individually test each prompt on the COREVQA dataset ourselves, to verify accuracy scores. The prompt achieving the highest performance wins!