Research map / Architectures and efficiency / RNNs and normalization
Non-Autoregressive Machine Translation with Auxiliary Regularization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add similarity and reconstruction regularization (ablation) | NAT-BASE · IWSLT14 De-En dev | BLEU: 28.73 -> 30.84 | Helped |
| Add target-token-conditioned similarity regularization (ablation) | NAT-BASE · IWSLT14 De-En dev | BLEU: 28.73 -> 30.02 | Helped |
| Add source-sentence reconstruction regularization (ablation) | NAT-BASE · IWSLT14 De-En dev | BLEU: 28.73 -> 30.21 | Helped |
| Universally penalize similarity between adjacent hidden states instead of conditioning it on target-token similarity (ablation) | NAT-BASE · IWSLT14 De-En dev | BLEU: 28.73 -> 28.32 | Hurt |
| Add de-duplication of repeated tokens (ablation) | NAT-BASE · IWSLT14 De-En dev | BLEU: 28.73 -> 29.45 | Helped |
| Add similarity and reconstruction regularization | NAT-BASE · WMT14 En-De | BLEU: NAT-REG (rescoring 9): $24.61$; LT (rescoring $100$): $22.5$ | Helped |
| Add similarity and reconstruction regularization | NAT-BASE · WMT14 De-En | BLEU: NAT-REG (rescoring 9): $28.90$; NAT-IR (10 refinements): $25.48$ | Helped |
| Add similarity and reconstruction regularization | NAT-BASE · IWSLT14 De-En | BLEU: NAT-REG (rescoring 9): $28.04$; NAT-IR (adaptive refinements): $24.63^\dagger$ | Helped |
| Add similarity and reconstruction regularization | NAT-BASE · IWSLT16 En-De | BLEU: NAT-REG (rescoring 9): $27.02$; NAT-FT (NPD rescoring $100$): $28.16$ | Hurt |
Closest papers by meaning
- Non-Autoregressive Neural Machine Translation with Enhanced Decoder Input (AAAI 2019 · similarity 0.93)
- Rephrasing the Reference for Non-autoregressive Machine Translation (AAAI 2023 · similarity 0.91)
- Guiding Non-Autoregressive Neural Machine Translation Decoding with Reordering Information (AAAI 2021 · similarity 0.90)
- Non-autoregressive Machine Translation with Probabilistic Context-free Grammar (NeurIPS 2023 · similarity 0.90)
- Non-Autoregressive Neural Machine Translation (ICLR 2018 · similarity 0.90)
- AEQA-NAT : Adaptive End-to-end Quantization Alignment Training Framework for Non-autoregressive Machine Translation (ICML 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.