What is the best recipe for character-level encoder-only modelling?

Kris Cao

This paper aims to benchmark recent progress in language understanding models that output contextualised representations at the character level. Many such modelling architectures a…