Machine Translation On Iwslt2014 German

BLEU score

评测结果

各个模型在此基准测试上的表现结果

比较表格

模型名称	BLEU score
attention-is-all-you-need	34.44
random-feature-attention-1	34.4
time-aware-large-kernel-convolutions	35.5
bi-simcut-a-simple-strategy-for-boosting-1	38.37
mask-attention-networks-rethinking-and	36.3
wide-minima-density-hypothesis-and-the	37.78
non-autoregressive-translation-by-learning	31.15
classical-structured-prediction-losses-for	32.84
bert-mbert-or-bibert-a-study-on	38.61
pay-less-attention-with-lightweight-and	34.8
autodropout-learning-dropout-patterns-to	35.8
pay-less-attention-with-lightweight-and	35.2
r-drop-regularized-dropout-for-neural	37.25
guidelines-for-the-regularization-of-gammas	35.1385
tag-less-back-translation	28.83
190506596	35.7
r-drop-regularized-dropout-for-neural	37.90
data-diversification-an-elegant-strategy-for	37.2
unidrop-a-simple-yet-effective-technique-to	36.88
an-actor-critic-algorithm-for-sequence	28.53
a-simple-but-tough-to-beat-data-augmentation	37.6
bi-simcut-a-simple-strategy-for-boosting-1	37.81
deterministic-reversible-data-augmentation	37.95
towards-neural-phrase-based-machine	30.08
latent-alignment-and-variational-attention	33.1
cipherdaug-ciphertext-based-data-augmentation	37.53
sequence-generation-with-mixed	36.41
muse-parallel-multi-scale-attention-for	36.3
relaxed-attention-for-transformer-models	37.96
multi-branch-attentive-transformer	36.22
rethinking-perturbations-in-encoder-decoders	36.22
integrating-pre-trained-language-model-into	40.43
autoregressive-knowledge-distillation-through	35.4
delight-very-deep-and-light-weight	35.3