The point is that they are comparing the English comprehension of various models, but their examples range from slightly incorrect to incomprehensible English.

The point is that they are comparing the English comprehension of various models, but their examples range from slightly incorrect to incomprehensible English. They don’t seem to be aware of this, because if you were intentionally testing on incorrect English you would also test on correct English as well to be able to quantify the difference. If I was writing a paper about AI comprehension of Italian and I was not confident at speaking Italian I would definitely want to ask an Italian speaker to check my examples for me.