List your company
ResearchFormerly WizWand

Machine Translation on WMT newstest 2016 (test)

40.4BLEUdir + lm + big
33.442435.248737.05538.8613Nov 13, 2020Feb 15, 2021May 20, 2021Aug 22, 2021Nov 24, 2021Feb 26, 2022Jun 1, 2022
Updated 8mo ago

Evaluation results

MethodLinks
dir + lm + bigmono tokens En (B)=4.5, mono tokens Ro (B)=22, language_model=true, channel_model_size=big, beam=50, backtranslation=true2020.1140.4
mBART02mono tokens En (B)=66, mono tokens Ro (B)=66, backtranslation=true2020.1139.9
mBART02mono tokens En (B)=66, mono tokens Ro (B)=66, backtranslation=true, beam=502020.1139.9
dir + lm + bigmono tokens En (B)=4.5, mono tokens Ro (B)=22, language_model=true, channel_model_size=big, backtranslation=true2020.1139.6
dir + lm + bigmono tokens En (B)=4.5, mono tokens Ro (B)=22, language_model=true, channel_model_size=big, beam=502020.1139.1
dir + lm + bigmono tokens En (B)=4.5, mono tokens Ro (B)=22, language_model=true, channel_model_size=big, beam=502020.1139.1
dir + lmmono tokens En (B)=4.5, mono tokens Ro (B)=22, language_model=true, backtranslation=true2020.1138.7
mBART02mono tokens En (B)=66, mono tokens Ro (B)=662020.1138.5
mBART02mono tokens En (B)=66, mono tokens Ro (B)=662020.1138.5
dirmono tokens En (B)=-, mono tokens Ro (B)=-, backtranslation=true2020.1138.4
dir + lm + bigmono tokens En (B)=4.5, mono tokens Ro (B)=22, language_model=true, channel_model_size=big2020.1137.7
dir + lm + bigmono tokens En (B)=4.5, mono tokens Ro (B)=22, language_model=true, channel_model_size=big2020.1137.3
dir + lmmono tokens En (B)=4.5, mono tokens Ro (B)=22, language_model=true2020.1135.9
dir + lmmono tokens En (B)=4.5, mono tokens Ro (B)=22, language_model=true2020.1135.4
B2T connectionEnc-Dec layers=18L-18L2022.0634.93
B2T connectionEnc-Dec layers=6L-6L2022.0634.73
T-FixupEnc-Dec layers=18L-18L2022.0634.69
dirmono tokens En (B)=-, mono tokens Ro (B)=-2020.1134.6
dirmono tokens En (B)=-, mono tokens Ro (B)=-2020.1134.6
T-FixupEnc-Dec layers=6L-6L2022.0634.56
Post-LNEnc-Dec layers=6L-6L2022.0634.4
DeepNetEnc-Dec layers=6L-6L2022.0634.39
DLCLEnc-Dec layers=6L-6L2022.0634.26
DeepNetEnc-Dec layers=18L-18L2022.0634.25
Pre-LNEnc-Dec layers=18L-18L2022.0634.16
RealFormerEnc-Dec layers=6L-6L2022.0634.15
AdminEnc-Dec layers=6L-6L2022.0634.12
DLCLEnc-Dec layers=18L-18L2022.0633.98
AdminEnc-Dec layers=18L-18L2022.0633.88
Pre-LNEnc-Dec layers=6L-6L2022.0633.84
RealFormerEnc-Dec layers=18L-18L2022.0633.71