mlsanityData to 14 Oct 2026

Deep Encoder, Shallow Decoder: Reevaluating Non-autoregressive Machine Translation

arXiv 2006.10369 ↗First posted 18 Jun 2020cs.CLcited by 105

Not on the map. Why?

Abstract (arXiv)

Much recent effort has been invested in non-autoregressive neural machine translation, which appears to be an efficient alternative to state-of-the-art autoregressive machine translation on modern GPUs. In contrast to the latter, where generation is sequential, the former allows generation to be parallelized across target token positions. Some of the latest non-autoregressive models have achieved impressive translation quality-speed tradeoffs compared to autoregressive baselines. In this work, we reexamine this tradeoff and argue that autoregressive baselines can be substantially sped up without loss in accuracy. Specifically, we study autoregressive models with encoders and decoders of varied depths. Our extensive experiments show that given a sufficiently deep encoder, a single-layer autoregressive decoder can substantially outperform strong non-autoregressive models with comparable inference speed. We show that the speed disadvantage for autoregressive baselines compared to non-autoregressive methods has been overestimated in three aspects: suboptimal layer allocation, insufficient speed measurement, and lack of knowledge distillation. Our results establish a new protocol for future research toward fast, accurate machine translation. Our code is available at https://github.com/jungokasai/deep-shallow.

Key facts. Cited by 105 papers in the map (data through 14 Oct 2026). First posted on arXiv 18 Jun 2020.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2021: 18; 2022: 39; 2023: 27; 2024: 6; 2025: 11; 2026: 4.

105 papers.

Sort:

By area:

  1. 2022Confident Adaptive Language Modelingcited by 310
  2. 2022Continuous diffusion for categorical datacited by 214
  3. 2022HyperTree Proof Search for Neural Theorem Provingcited by 197
  4. 2022Speculative Decoding: Exploiting Speculative Execution for Accelerating Seq2seq Generationcited by 171
  5. 2023Speculative Decoding with Big Little Decodercited by 164
  6. 2021Charformer: Fast Character Transformers via Gradient-based Subword Tokenizationcited by 160
  7. 2022Scene Text Recognition with Permuted Autoregressive Sequence Modelscited by 149
  8. 2023Accelerating Transformer Inference for Translation via Parallel Decodingcited by 124
  9. 2021ImageBART: Bidirectional Context with Multinomial Diffusion for Autoregressive Image Synthesiscited by 122
  10. 2023f-Divergence Minimization for Sequence-Level Knowledge Distillationcited by 114
  11. 2021Scaling Laws for Neural Machine Translationcited by 111
  12. 2023Towards Efficient Generative Large Language Model Serving: A Survey from Algorithms to Systemscited by 102
38 more
  1. 2021Finetuning Pretrained Transformers into RNNscited by 96
  2. 2023Batch Prompting: Efficient Inference with Large Language Model APIscited by 85
  3. 2021CPT: A Pre-Trained Unbalanced Transformer for Both Chinese Language Understanding and Generationcited by 81
  4. 2021Scalable and Efficient MoE Training for Multitask Multilingual Modelscited by 76
  5. 2022A Survey on Non-Autoregressive Generation for Neural Machine Translation and Beyondcited by 64
  6. 2022UnitY: Two-pass Direct Speech-to-speech Translation with Discrete Unitscited by 60
  7. 2021Order-Agnostic Cross Entropy for Non-Autoregressive Machine Translationcited by 57
  8. 2022Data Scaling Laws in NMT: The Effect of Noise and Architecturecited by 55
  9. 2023LLMCad: Fast and Scalable On-device Large Language Model Inferencecited by 45
  10. 2022Directed Acyclic Transformer for Non-Autoregressive Machine Translationcited by 44
  11. 2022DiffusER: Discrete Diffusion via Edit-based Reconstructioncited by 41
  12. 2021Non-Autoregressive Translation with Layer-Wise Prediction and Deep Supervisioncited by 40
  13. 2023Predictive Pipelined Decoding: A Compute-Latency Trade-off for Exact LLM Decodingcited by 40
  14. 2023Lookahead: An Inference Acceleration Framework for Large Language Model with Lossless Generation Accuracycited by 37
  15. 2021ABC: Attention with Bounded-memory Controlcited by 36
  16. 2023Mixture of Quantized Experts (MoQE): Complementary Effect of Low-bit Quantization and Robustnesscited by 34
  17. 2024Decoding Speculative Decodingcited by 34
  18. 2021Bidimensional Leaderboards: Generate and Evaluate Language Hand in Handcited by 32
  19. 2022Linear Complexity Randomized Self-attention Mechanismcited by 31
  20. 2023UniTabE: A Universal Pretraining Protocol for Tabular Foundation Model in Data Sciencecited by 29
  21. 2025Encoder-Decoder Gemma: Improving the Quality-Efficiency Trade-Off via Adaptationcited by 28
  22. 2022On the Learning of Non-Autoregressive Transformerscited by 27
  23. 2022Who Says Elephants Can't Run: Bringing Large Scale MoE Models into Cloud Scale Productioncited by 27
  24. 2022Examining Scaling and Transfer of Language Model Architectures for Machine Translationcited by 26
  25. 2021Diversifying Dialog Generation via Adaptive Label Smoothingcited by 25
  26. 2025Latent Thought Models with Variational Bayes Inference-Time Computationcited by 25
  27. 2021AligNART: Non-autoregressive Neural Machine Translation by Jointly Learning to Estimate Alignment and Translatecited by 24
  28. 2023Multilingual Pixel Representations for Translation and Effective Cross-lingual Transfercited by 22
  29. 2022EdgeFormer: A Parameter-Efficient Transformer for On-Device Seq2seq Generationcited by 21
  30. 2021Sentence Bottleneck Autoencoders from Transformer Language Modelscited by 19
  31. 2022E2S2: Encoding-Enhanced Sequence-to-Sequence Pretraining for Language Understanding and Generationcited by 19
  32. 2022Non-Monotonic Latent Alignments for CTC-Based Non-Autoregressive Machine Translationcited by 16
  33. 2022SMaLL-100: Introducing Shallow Multilingual Machine Translation Model for Low-Resource Languagescited by 16
  34. 2023One Wide Feedforward is All You Needcited by 15
  35. 2022Viterbi Decoding of Directed Acyclic Transformer for Non-Autoregressive Machine Translationcited by 13
  36. 2023Learning Language-Specific Layers for Multilingual Machine Translationcited by 13
  37. 2022Non-Autoregressive Machine Translation: It's Not as Fast as it Seemscited by 11
  38. 2023Fuzzy Alignments in Directed Acyclic Graph for Non-Autoregressive Machine Translationcited by 11

50 most cited of 105 All 105 (JSON)