Joe Welson

First of all, thank you for your work. I found that the evaluation for the multi-label dataset I(e.g., IMDB) is unreasonable and unjustified, which leads your method to the state-of-the-art...