About train/test split of RCV1

[yzhangcs]

Hi, thanks for your nice paper and code!
I have noticed that the standard split for RCV1 train/test in the original paper is 23,149/781,265. But from the data downloaded from your link, I found the train file is much bigger than test.
I wonder is this correct? Thanks in advance.