Research map / Architectures and efficiency / RNNs and normalization
Fast Structured Decoding for Sequence Models
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 CRF structured decoding to NART | NART · WMT14 En-De | BLEU: 20.27 -> 23.32 | Helped |
| add CRF structured decoding to NART | NART · WMT14 De-En | BLEU: 22.02 -> 25.75 | Helped |
| add CRF structured decoding to NART | NART · IWSLT14 De-En | BLEU: 23.04 -> 26.39 | Helped |
| add position-dependent dynamic transitions (ablation) | NART-CRF · WMT14 En-De | BLEU: +0.12 BLEU | Helped |
| add position-dependent dynamic transitions (ablation) | NART-CRF · WMT14 De-En | BLEU: +1.47 BLEU | Helped |
| add position-dependent dynamic transitions (ablation) | NART-CRF · IWSLT14 De-En | BLEU: +1.05 BLEU | Helped |
| increase CRF beam size from 1 to 16 (ablation) | NART-CRF · WMT14 En-De | BLEU: 15.10 -> 23.22 | Helped |
| increase CRF beam size beyond 16 (ablation) | NART-CRF · WMT14 En-De | BLEU: 23.22 -> 23.38 | No clear effect |
| increase CRF beam size from 1 to 16 with rescoring 9 (ablation) | NART-CRF (rescoring 9) · WMT14 En-De | BLEU: 19.61 -> 25.93 | Helped |
| increase CRF beam size beyond 16 with rescoring 9 (ablation) | NART-CRF (rescoring 9) · WMT14 En-De | BLEU: 25.93 -> 26.08 | No clear effect |
| increase CRF beam size from 1 to 16 with rescoring 19 (ablation) | NART-CRF (rescoring 19) · WMT14 En-De | BLEU: 20.02 -> 26.57 | Helped |
| increase CRF beam size beyond 16 with rescoring 19 (ablation) | NART-CRF (rescoring 19) · WMT14 En-De | BLEU: 26.57 -> 26.66 | No clear effect |
Closest papers by meaning
- Latent-Variable Non-Autoregressive Neural Machine Translation with Deterministic Inference Using a Delta Posterior (AAAI 2020 · similarity 0.91)
- Deep Encoder, Shallow Decoder: Reevaluating Non-autoregressive Machine Translation (ICLR 2021 · similarity 0.91)
- Non-Autoregressive Neural Machine Translation (ICLR 2018 · similarity 0.91)
- When Linear Attention Meets Autoregressive Decoding: Towards More Effective and Efficient Linearized Large Language Models (ICML 2024 · similarity 0.88)
- Flexible Language Modeling in Continuous Space with Transformer-based Autoregressive Flows (NeurIPS 2025 · similarity 0.88)
- Fast and Fluent Diffusion Language Models via Convolutional Decoding and Rejective Fine-tuning (NeurIPS 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.