What Have We Achieved on Non-autoregressive Translation?
Not on the map. Why?
Abstract (arXiv)
Recent advances have made non-autoregressive (NAT) translation comparable to autoregressive methods (AT). However, their evaluation using BLEU has been shown to weakly correlate with human annotations. Limited research compares non-autoregressive translation and autoregressive translation comprehensively, leaving uncertainty about the true proximity of NAT to AT. To address this gap, we systematically evaluate four representative NAT methods across various dimensions, including human evaluation. Our empirical results demonstrate that despite narrowing the performance gap, state-of-the-art NAT still underperforms AT under more reliable evaluation metrics. Furthermore, we discover that explicitly modeling dependencies is crucial for generating natural language and generalizing to out-of-distribution sequences.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 42 of its 54 references that are in the map. First posted on arXiv 21 May 2024.
Builds on
42 of 54 references.
References · 42
- 2017Attention Is All You Needcited by 60,361
- 2014Adam: A Method for Stochastic Optimizationcited by 37,240
- 2015Neural Machine Translation of Rare Words with Subword Unitscited by 2,918
- 2020BLEURT: Learning Robust Metrics for Text Generationcited by 1,203
- 2020COMET: A Neural Framework for MT Evaluationcited by 1,033
- 2015Scheduled Sampling for Sequence Prediction with Recurrent Neural Networkscited by 803
- 2015Sequence Level Training with Recurrent Neural Networkscited by 576
- 2017Non-Autoregressive Neural Machine Translationcited by 463
- 2023Large Language Models Are State-of-the-Art Evaluators of Translation Qualitycited by 420
- 2021Experts, Errors, and Context: A Large-Scale Study of Human Evaluation for Machine Translationcited by 338
- 2019Levenshtein Transformercited by 199
- 2019Understanding Knowledge Distillation in Non-autoregressive Machine Translationcited by 122
30 more
- 2018Fast Decoding in Sequence Models using Discrete Latent Variablescited by 104
- 2020Glancing Transformer for Non-Autoregressive Neural Machine Translationcited by 103
- 2020Non-Autoregressive Machine Translation with Latent Alignmentscited by 96
- 2021Learning Compact Metrics for MTcited by 93
- 2019FlowSeq: Non-Autoregressive Conditional Sequence Generation with Generative Flowcited by 89
- 2018End-to-End Non-Autoregressive Neural Machine Translation with Connectionist Temporal Classificationcited by 72
- 2019Domain Robustness in Neural Machine Translationcited by 65
- 2020Aligned Cross Entropy for Non-Autoregressive Machine Translationcited by 65
- 2022A Survey on Non-Autoregressive Generation for Neural Machine Translation and Beyondcited by 64
- 2019Fast Structured Decoding for Sequence Modelscited by 59
- 2021Order-Agnostic Cross Entropy for Non-Autoregressive Machine Translationcited by 57
- 2020Understanding and Improving Lexical Choice in Non-Autoregressive Translationcited by 56
- 2022Towards Opening the Black Box of Neural Machine Translation: Source and Target Interpretations of the Transformercited by 45
- 2022Directed Acyclic Transformer for Non-Autoregressive Machine Translationcited by 44
- 2023Understanding and Detecting Hallucinations in Neural Machine Translation via Model Introspectioncited by 44
- 2019Minimizing the Bag-of-Ngrams Difference for Non-Autoregressive Neural Machine Translationcited by 40
- 2020On the Importance of Word and Sentence Representation Learning in Implicit Discourse Relation Classificationcited by 40
- 2019Fine-Tuning by Curriculum Learning for Non-Autoregressive Neural Machine Translationcited by 39
- 2020Evaluating Robustness to Input Perturbations for Neural Machine Translationcited by 36
- 2021On Compositional Generalization of Neural Machine Translationcited by 34
- 2021Rejuvenating Low-Frequency Words: Making the Most of Parallel Data in Non-Autoregressive Translationcited by 34
- 2022On the Learning of Non-Autoregressive Transformerscited by 27
- 2021Non-Autoregressive Translation by Learning Target Categorical Codescited by 25
- 2021Progressive Multi-Granularity Training for Non-Autoregressive Translationcited by 25
- 2021Language Modeling, Lexical Translation, Reordering: The Training Process of NMT through the Lens of Classical SMTcited by 21
- 2022textitlatent-GLAT: Glancing at Latent Variables for Parallel Text Generationcited by 20
- 2019Retrieving Sequential Information for Non-Autoregressive Neural Machine Translationcited by 15
- 2022Multi-Granularity Optimization for Non-Autoregressive Translationcited by 6
- 2022Non-Autoregressive Neural Machine Translation: A Call for Claritycited by 5
- 2021Digging Errors in NMT: Evaluating and Understanding Model Errors from Partial Hypothesis Spacecited by 2
Built on by
Citing papers per year: 2026: 1.
1 paper.