Research map / Probabilistic and scientific modeling / Variational and generative sampling
Latent-Variable Non-Autoregressive Neural Machine Translation with Deterministic Inference Using a Delta Posterior
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 deterministic inference with one refinement step (ablation) | LaNMT with knowledge distillation · ASPEC Ja-En | BLEU: 25.2 -> 27.5 | Helped |
| Add deterministic inference with one refinement step (ablation) | LaNMT with knowledge distillation · WMT'14 En-De | BLEU: 22.2 -> 24.1 | Helped |
| Add sequence-level knowledge distillation (ablation) | LaNMT · ASPEC Ja-En | BLEU: 13.3 -> 25.2 | Helped |
| Add sequence-level knowledge distillation (ablation) | LaNMT · WMT'14 En-De | BLEU: 11.8 -> 22.2 | Helped |
| Search multiple initial latent variables and rescore candidates with the teacher (ablation) | LaNMT with deterministic inference · ASPEC Ja-En | BLEU: 27.5 -> 28.3 | Helped |
| Search multiple initial latent variables and rescore candidates with the teacher (ablation) | LaNMT with deterministic inference · WMT'14 En-De | BLEU: 24.1 -> 25.1 | Helped |
| Run additional deterministic refinement steps beyond the first (ablation) | LaNMT · ASPEC Ja-En and WMT'14 En-De | BLEU | No clear effect |
| Vary latent dimensionality within 4 to 32 (ablation) | LaNMT · ASPEC Ja-En and WMT'14 En-De | BLEU | No clear effect |
Closest papers by meaning
- Non-Autoregressive Neural Machine Translation (ICLR 2018 · similarity 0.93)
- Fast Structured Decoding for Sequence Models (NeurIPS 2019 · similarity 0.91)
- Minimizing the Bag-of-Ngrams Difference for Non-Autoregressive Neural Machine Translation (AAAI 2020 · similarity 0.90)
- Fast Decoding in Sequence Models Using Discrete Latent Variables (ICML 2018 · similarity 0.90)
- Calibrating Translation Decoding with Quality Estimation on LLMs (NeurIPS 2025 · similarity 0.87)
- AEQA-NAT : Adaptive End-to-end Quantization Alignment Training Framework for Non-autoregressive Machine Translation (ICML 2025 · 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.