search

ictnlp / PLUVR

Code for ACL 2022 main conference paper "Neural Machine Translation with Phrase-Level Universal Visual Representations".
21 stars 6 forks source link

BLEU with UVR-NMT #5

Closed Hannibal046 closed 2 years ago

Hannibal046 commented 2 years ago

Hi, thanks for the great work. I am wondering why there is an obvious gap between the BLEU score in your paper and that in UVR-NMT, since you all conduct experiments on Multi30k dataset. Do I miss something here ?

Your paper

2203 10299 pdf 2022-06-30 16-46-39

UVR-NMT paper

pdf?id=Byl8hhNYPS 2022-06-30 16-48-18
Poeroz commented 2 years ago

Hi, thanks for your interest in our paper! We implement their method with a stronger baseline (attention heads=4, dropout=0.3), while they conducted experiments with Transformer-base, so our BLEU scores are higher than theirs.