Enquiry about the paper

[riven314]

Thanks for the great work! The benchmark mainly use style of binary class (e.g. positive v.s. negative). Could the model work on style of multiple classes? (e.g. write style 1, write style 2, ... ,write style N)