AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Duplex Sequence-to-Sequence Learning for Reversible Machine Translation

Zaixiang Zheng, Hao Zhou, Shujian Huang, Jiajun Chen, Jingjing Xu, Lei Li

NeurIPS 2021 · Architectures and efficiency › RNNs and normalization · approach: Non-autoregressive translation · 10 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (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 changedWhereMeasuredEffect
train duplex (bidirectional with reversible model) instead of simplex single direction (ablation)REDER · WMT14 En-DeBLEU: 26.20 -> 27.50Improved
use multitask learning for bidirectional translation instead of duplex (ablation)REDER · WMT14 En-DeBLEU: 26.20 -> 25.58Hurt
add multitask learning to AT Transformer-baseTransformer-base · WMT14 En-DeBLEU: 27.60 -> 27.06Hurt
add back-translation without monolingual dataAT Transformer-base · WMT14 En-DeBLEU: 27.60 -> 27.82Improved
add knowledge distillation (ablation)NAT without CTC · WMT14 En-DeBLEU: 11.40 -> 19.50Improved
add CTC (ablation)KD NAT · WMT14 En-DeBLEU: 19.50 -> 25.01Improved
add relative self attention (ablation)KD+CTC REDER · WMT14 En-DeBLEU: 25.01 -> 25.55Improved
add reversible architecture (revnet) (ablation)KD+CTC+R-SAN · WMT14 En-DeBLEU: 25.55 -> 26.20Improved
add forward-backward agreement and cycle consistency losses (ablation)REDER · WMT14 En-DeBLEU: 26.20 -> 26.89Improved
add AT reranking to beam search decoding (ablation)REDER · WMT14 En-DeBLEU: 26.90 -> 27.50Improved
use CTC beam search instead of greedy decoding (ablation)REDER · WMT14 En-DeBLEU: 26.89 -> 26.90No clear effect
train on separate per-direction KD data instead of raw data (ablation)REDER · WMT14 En-DeBLEU: 17.85 -> 27.50Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy