Extremely Small BERT Models from Mixed-Vocabulary Training.

Sanqiang Zhao, Raghav Gupta, Yang Song, Denny Zhou

Browse the full EACL paper archive.