List your company
ResearchFormerly WizWand

Machine Translation on IWSLT De-En combined 2014 (test)

37.2BLEU ScoreOur Data Diversification with Dynamic Conv
27.63230.11632.635.084Nov 5, 2019
Updated 8mo ago

Evaluation results

MethodLinks
Our Data Diversification with Dynamic ConvModel Architecture=Dynamic Conv, Training Strategy=Data Diversification2019.1137.2
Our Data Diversification with TransformerModel Architecture=Transformer, Training Strategy=Data Diversification2019.1137
Dynamic ConvModel Architecture=Dynamic Conv, Training Strategy=Baseline2019.1135
TransformerModel Architecture=Transformer, Training Strategy=Baseline2019.1134.7
Transformer with Multi-AgentModel Architecture=Transformer, Training Strategy=Multi-Agent2019.1134.7
Transformer with Ens-DistillModel Architecture=Transformer, Training Strategy=Ensemble Distillation2019.1134.7
Transformer with Distill (T=S)Model Architecture=Transformer, Training Strategy=Knowledge Distillation (T=S)2019.1134.1
Transformer with Distill (T>S)Model Architecture=Transformer, Training Strategy=Knowledge Distillation (T>S)2019.1133.6
Our Data Diversification with TransformerModel Architecture=Transformer, Training Strategy=Data Diversification2019.1130.6
Our Data Diversification with Dynamic ConvModel Architecture=Dynamic Conv, Training Strategy=Data Diversification2019.1130.6
Transformer with Multi-AgentModel Architecture=Transformer, Training Strategy=Multi-Agent2019.1128.9
Transformer with Ens-DistillModel Architecture=Transformer, Training Strategy=Ensemble Distillation2019.1128.8
Dynamic ConvModel Architecture=Dynamic Conv, Training Strategy=Baseline2019.1128.7
TransformerModel Architecture=Transformer, Training Strategy=Baseline2019.1128.6
Transformer with Distill (T=S)Model Architecture=Transformer, Training Strategy=Knowledge Distillation (T=S)2019.1128.5
Transformer with Distill (T>S)Model Architecture=Transformer, Training Strategy=Knowledge Distillation (T>S)2019.1128