Has there been a comparative evaluation on the performance improvement differences between having a bbox and not having a bbox in question and answer?

[zhongtao93]

For example：
If delete all coords in question and answer, how much the performance will decrease?
ori answer:
Two shirtless men[0.284, 0.446, 0.400, 0.582][0.222, 0.480, 0.314, 0.586] are standing on a balcony[0.118, 0.586, 0.526, 0.786] in front of an open window[0.112, 0.564, 0.532, 0.764]. They are shielding their eyes and pointing at the sky, observing something interesting. It seems that they are on an older building[0.000, 0.166, 0.998, 0.832].
del coords answer:
Two shirtless men are standing on a balcony in front of an open window. They are shielding their eyes and pointing at the sky, observing something interesting. It seems that they are on an older building.