Research map / Probabilistic and scientific modeling / Molecules and proteins
Non-autoregressive Machine Translation with Probabilistic Context-free Grammar
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 |
|---|---|---|---|
| Model target-side dependencies with a Right-Heavy PCFG | NAT · WMT14, WMT17, and WMT16 translation benchmarks | BLEU | Helped |
| Use Viterbi decoding instead of greedy decoding (ablation) | PCFG-NAT · WMT14 En-De | BLEU: 26.01 -> 27.02 | Helped |
| Use Viterbi decoding instead of greedy decoding (ablation) | PCFG-NAT · WMT14 De-En | BLEU: 31.38 -> 32.29 | Helped |
| Use Viterbi decoding instead of greedy decoding (ablation) | PCFG-NAT · WMT17 En-Zh | BLEU: 32.82 -> 33.60 | Helped |
| Use Viterbi decoding instead of greedy decoding (ablation) | PCFG-NAT · WMT17 Zh-En | BLEU: 22.50 -> 23.40 | Helped |
| Use Viterbi decoding instead of greedy decoding (ablation) | PCFG-NAT · WMT16 En-Ro | BLEU: 32.05 -> 32.72 | Helped |
| Use Viterbi decoding instead of greedy decoding (ablation) | PCFG-NAT · WMT16 Ro-En | BLEU: 32.65 -> 33.07 | Helped |
| Use Viterbi rather than greedy decoding (ablation) | PCFG-NAT · WMT14 En-De test | speedup: 1.00× -> 0.91× | Hurt |
| Add GLAT (ablation) | PCFG-NAT · WMT16 En-Ro | BLEU: 31.94 -> 32.72 | Helped |
| Add knowledge distillation after GLAT (ablation) | PCFG-NAT · WMT16 En-Ro | BLEU: 32.72 -> 32.75 | Helped |
| Add GLAT (ablation) | DA Transformer · WMT16 En-Ro | BLEU: 31.98 -> 32.46 | Helped |
| Add knowledge distillation after GLAT (ablation) | DA Transformer · WMT16 En-Ro | BLEU: 32.46 -> 32.40 | Hurt |
Closest papers by meaning
- Non-Autoregressive Machine Translation with Auxiliary Regularization (AAAI 2019 · similarity 0.90)
- Improving Non-Autoregressive Translation Models Without Distillation (ICLR 2022 · similarity 0.90)
- RenewNAT: Renewing Potential Translation for Non-autoregressive Transformer (AAAI 2023 · similarity 0.90)
- AEQA-NAT : Adaptive End-to-end Quantization Alignment Training Framework for Non-autoregressive Machine Translation (ICML 2025 · similarity 0.88)
- Flexible Language Modeling in Continuous Space with Transformer-based Autoregressive Flows (NeurIPS 2025 · similarity 0.86)
- DiffNorm: Self-Supervised Normalization for Non-autoregressive Speech-to-speech Translation (NeurIPS 2024 · similarity 0.85)
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.