Machine Translation On Iwslt2014 German

评估指标

BLEU score

评测结果

各个模型在此基准测试上的表现结果

		Paper Title	Repository
PiNMT	40.43	Integrating Pre-trained Language Model into Neural Machine Translation	-
BiBERT	38.61	BERT, mBERT, or BiBERT? A Study on Contextualized Embeddings for Neural Machine Translation
Bi-SimCut	38.37	Bi-SimCut: A Simple Strategy for Boosting Neural Machine Translation
Cutoff + Relaxed Attention + LM	37.96	Relaxed Attention for Transformer Models
DRDA	37.95	Deterministic Reversible Data Augmentation for Neural Machine Translation
Transformer + R-Drop + Cutoff	37.90	R-Drop: Regularized Dropout for Neural Networks
SimCut	37.81	Bi-SimCut: A Simple Strategy for Boosting Neural Machine Translation
Cutoff+Knee	37.78	Wide-minima Density Hypothesis and the Explore-Exploit Learning Rate Schedule
Cutoff	37.6	A Simple but Tough-to-Beat Data Augmentation Approach for Natural Language Understanding and Generation
CipherDAug	37.53	CipherDAug: Ciphertext based Data Augmentation for Neural Machine Translation
Transformer + R-Drop	37.25	R-Drop: Regularized Dropout for Neural Networks
Data Diversification	37.2	Data Diversification: A Simple Strategy For Neural Machine Translation
UniDrop	36.88	UniDrop: A Simple yet Effective Technique to Improve Transformer without Extra Cost	-
MixedRepresentations	36.41	Sequence Generation with Mixed Representations	-
Mask Attention Network (small)	36.3	Mask Attention Networks: Rethinking and Strengthen Transformer
MUSE(Parallel Multi-scale Attention)	36.3	MUSE: Parallel Multi-Scale Attention for Sequence to Sequence Learning
MAT	36.22	Multi-branch Attentive Transformer
Transformer+Rep(Sim)+WDrop	36.22	Rethinking Perturbations in Encoder-Decoders for Fast Training
TransformerBase + AutoDropout	35.8	AutoDropout: Learning Dropout Patterns to Regularize Deep Networks
Local Joint Self-attention	35.7	Joint Source-Target Self Attention with Locality Constraints

0 of 34 row(s) selected.