Zero-shot off-target translation in multilingual NMT is caused by a learned shortcut toward the central language, and removing such examples late in training lets models forget it.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
On the Shortcut Learning in Multilingual Neural Machine Translation
Zero-shot off-target translation in multilingual NMT is caused by a learned shortcut toward the central language, and removing such examples late in training lets models forget it.