FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flow

5 Sep 2019Xuezhe MaChunting ZhouXian LiGraham NeubigEduard Hovy

Most sequence-to-sequence (seq2seq) models are autoregressive; they generate each token by conditioning on previously generated tokens. In contrast, non-autoregressive seq2seq models generate all tokens in one pass, which leads to increased efficiency through parallel processing on hardware such as GPUs... (read more)

PDF Abstract

Evaluation results from the paper


Task Dataset Model Metric name Metric value Global rank Compare
Machine Translation IWSLT2015 German-English FlowSeq-base BLEU score 24.75 # 15
Machine Translation WMT2014 English-German FlowSeq-large (NPD n = 30) BLEU score 23.64 # 21
Machine Translation WMT2014 English-German FlowSeq-base BLEU score 18.55 # 29
Machine Translation WMT2014 English-German FlowSeq-large BLEU score 20.85 # 25
Machine Translation WMT2014 English-German FlowSeq-large (IWD n = 15) BLEU score 22.94 # 23
Machine Translation WMT2014 English-German FlowSeq-large (NPD n = 15) BLEU score 23.14 # 22
Machine Translation WMT2014 German-English FlowSeq-large (NPD n = 15) BLEU score 27.71 # 2
Machine Translation WMT2014 German-English FlowSeq-base BLEU score 23.36 # 6
Machine Translation WMT2014 German-English FlowSeq-large BLEU score 25.40 # 5
Machine Translation WMT2014 German-English FlowSeq-large (IWD n = 15) BLEU score 27.16 # 3
Machine Translation WMT2014 German-English FlowSeq-large (NPD n = 30) BLEU score 28.29 # 1
Machine Translation WMT2016 English-Romanian FlowSeq-base BLEU score 29.26 # 8
Machine Translation WMT2016 English-Romanian FlowSeq-large BLEU score 29.86 # 5
Machine Translation WMT2016 English-Romanian FlowSeq-large (NPD n = 30) BLEU score 32.35 # 1
Machine Translation WMT2016 English-Romanian FlowSeq-large (NPD n = 15) BLEU score 31.97 # 2
Machine Translation WMT2016 English-Romanian FlowSeq-large (IWD n = 15) BLEU score 31.08 # 3
Machine Translation WMT2016 Romanian-English FlowSeq-large (NPD n = 30) BLEU score 32.91 # 4
Machine Translation WMT2016 Romanian-English FlowSeq-large (IWD n = 15) BLEU score 32.03 # 6
Machine Translation WMT2016 Romanian-English FlowSeq-base BLEU score 30.16 # 10
Machine Translation WMT2016 Romanian-English FlowSeq-large BLEU score 30.69 # 8
Machine Translation WMT2016 Romanian-English FlowSeq-large (NPD n = 15) BLEU score 32.46 # 5