A training method that distills soft targets from the current model's own word embedding distances yields small reported gains in language modeling and neural machine translation.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Self-Knowledge Distillation in Natural Language Processing
A training method that distills soft targets from the current model's own word embedding distances yields small reported gains in language modeling and neural machine translation.