Research map / Architectures and efficiency / RNNs and normalization
Rephrasing the Reference for Non-autoregressive Machine Translation
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 |
|---|---|---|---|
| Rephrase references according to NAT output | Vanilla NAT · WMT14 En-De | BLEU: 20.42 -> 25.33 | Helped |
| Rephrase references according to NAT output | CMLM · WMT14 En-De | BLEU: 18.21 -> 26.65 | Helped |
| Rephrase references according to NAT output | CTC · WMT14 En-De | BLEU: 26.34 -> 27.32 | Helped |
| Add a rephraser (ablation) | CTC · WMT14 En-De validation | BLEU: 24.77 -> 25.54 | Helped |
| Add 2 decoder layers without a rephraser (ablation) | CTC · WMT14 En-De validation | BLEU: 24.77 -> 24.89 | No clear effect |
| Add 30K training steps without a rephraser (ablation) | CTC · WMT14 En-De validation | BLEU: 24.77 -> 24.80 | No clear effect |
| Add 2 decoder layers and 30K steps without a rephraser (ablation) | CTC · WMT14 En-De validation | BLEU: 24.77 -> 24.87 | No clear effect |
| Swap the rephraser inputs (ablation) | Vanilla NAT with rephraser · WMT14 En-De validation | BLEU: 24.06 -> 24.01 | No clear effect |
| Use a Transformer encoder on fused inputs instead of the decoder design (ablation) | Vanilla NAT with rephraser · WMT14 En-De validation | BLEU: 24.06 -> 23.61 | Hurt |
| Reduce rephraser depth from 2 to 1 (ablation) | Vanilla NAT with rephraser · WMT14 En-De validation | BLEU: 24.06 -> 23.53 | Hurt |
| Increase rephraser depth from 2 to 4 (ablation) | Vanilla NAT with rephraser · WMT14 En-De validation | BLEU: 24.06 -> 24.09 | No clear effect |
| Use METEOR instead of BLEU for similarity reward (ablation) | Vanilla NAT with rephraser · WMT14 En-De validation | BLEU: 24.06 -> 23.71 | Hurt |
Closest papers by meaning
- Non-Autoregressive Neural Machine Translation (ICLR 2018 · similarity 0.91)
- Non-Autoregressive Machine Translation with Auxiliary Regularization (AAAI 2019 · similarity 0.91)
- Non-Autoregressive Neural Machine Translation with Enhanced Decoder Input (AAAI 2019 · similarity 0.91)
- Minimizing the Bag-of-Ngrams Difference for Non-Autoregressive Neural Machine Translation (AAAI 2020 · similarity 0.90)
- AEQA-NAT : Adaptive End-to-end Quantization Alignment Training Framework for Non-autoregressive Machine Translation (ICML 2025 · similarity 0.88)
- DiffNorm: Self-Supervised Normalization for Non-autoregressive Speech-to-speech Translation (NeurIPS 2024 · similarity 0.87)
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.