For decades people have been complaining about BLEU score not being a good metric for Machine Translation, and yet it is the metric that helped us identify the Transformer .
Haters gon' hate

→ View original post on X — @reza_zadeh,