Yes, totally agree. And wait what, the data set isn't public? The data set isn't public and we get to see only handpicked examples in the paper?