FlowSeq-base

Reported on 5 benchmarks across 1 task · 1 paper

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Natural Language Processing5 results

Machine TranslationonIWSLT2015 German-English
BLEU score· 2019-09-05
24.75
best: 36.2 (PS-KD)
FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow arXiv:1909.02480
Machine TranslationonWMT2016 Romanian-English
BLEU score· 2019-09-05
30.16
best: 40.3 (fast-noisy-channel-modeling)
FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow arXiv:1909.02480
Machine TranslationonWMT2014 German-English
BLEU score· 2019-09-05
23.36
best: 35.15 (Bi-SimCut)
FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow arXiv:1909.02480
Machine TranslationonWMT2014 English-German
BLEU score· 2019-09-05
18.55
best: 35.14 (Transformer Cycle (Rev))
FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow arXiv:1909.02480
Machine TranslationonWMT2016 English-Romanian
BLEU score· 2019-09-05
29.26
best: 34.7 (DeLighT)
FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow arXiv:1909.02480