Machine Translation On Wmt2016 Romanian

المقاييس

BLEU score

النتائج

نتائج أداء النماذج المختلفة على هذا المعيار القياسي

		Paper Title	Repository
fast-noisy-channel-modeling	40.3	Language Models not just for Pre-training: Fast Online Neural Noisy Channel Modeling
FLAN 137B (few-shot, k=9)	38.1	Finetuned Language Models Are Zero-Shot Learners
FLAN 137B (zero-shot)	37.3	Finetuned Language Models Are Zero-Shot Learners
MLM pretraining	35.3	Cross-lingual Language Model Pretraining
GenTranslate	33.5	GenTranslate: Large Language Models are Generative Multilingual Speech and Machine Translators
Attentional encoder-decoder + BPE	33.3	Edinburgh Neural Machine Translation Systems for WMT 16
Levenshtein Transformer (distillation)	33.26	Levenshtein Transformer
CMLM+LAT+4 iterations	33.26	Incorporating a Local Translation Mechanism into Non-autoregressive Translation
Adaptively Sparse Transformer (1.5-entmax)	33.1	Adaptively Sparse Transformers
HeadMask (Impt-18)	32.95	Alleviating the Inequality of Attention Heads for Neural Machine Translation	-
FlowSeq-large (NPD n = 30)	32.91	FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow
Adaptively Sparse Transformer (alpha-entmax)	32.89	Adaptively Sparse Transformers
HeadMask (Random-18)	32.85	Alleviating the Inequality of Attention Heads for Neural Machine Translation	-
FlowSeq-large (NPD n = 15)	32.46	FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow
FlowSeq-large (IWD n = 15)	32.03	FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow
NAT +FT + NPD	31.44	Non-Autoregressive Neural Machine Translation
CMLM+LAT+1 iterations	31.24	Incorporating a Local Translation Mechanism into Non-autoregressive Translation
FlowSeq-large	30.69	FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow
Denoising autoencoders (non-autoregressive)	30.30	Deterministic Non-Autoregressive Neural Sequence Modeling by Iterative Refinement
FlowSeq-base	30.16	FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow

0 of 21 row(s) selected.

Command Palette

Machine Translation On Wmt2016 Romanian

المقاييس

النتائج