Research map / Label-efficient and robust learning / Recognition and long-tail
Transductive Ensemble Learning for Neural 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 |
|---|---|---|---|
| TEL instead of token-level ensemble, strong WMT+LSN models | TEL · WMT En->De newstest2016 | BLEU: 41.2 -> 41.5 | Helped |
| TEL instead of ensemble, WMT+BT De->En newstest2016 | TEL · WMT De->En newstest2016 | BLEU: 44.3 -> 44.90 | Helped |
| TEL instead of ensemble, supervised WMT En->De | TEL · WMT En->De newstest2016 | BLEU: 36.2 -> 36.6 ± 0.09 | Helped |
| TEL instead of ensemble, supervised De->En newstest2017 | TEL · WMT De->En newstest2017 | BLEU: 35.6 -> 35.5 ± 0.10 | No clear effect |
| use more trajectories m with TEL (ablation) | TEL · WMT En->De newstest2016 | BLEU | Helped |
| use 4 trajectories x 5 models vs 1 trajectory x 20 models (ablation) | TEL · WMT En->De newstest2016 | BLEU | Helped |
| aggregate K>=5 models in direct ensemble (not TEL) (ablation) | TEL · WMT En->De newstest2016 | BLEU | Hurt |
| include validation and target test source sentences for TEL fine-tuning (ablation) | TEL · WMT En->De newstest2016 | BLEU: 34.6 -> 36.9 | Helped |
| use only validation set (newstest15) source for TEL (ablation) | TEL · WMT En->De newstest2016 | BLEU: 34.6 -> 35.4 | Helped |
| vary training/test data composition for TEL fine-tuning (ablation) | TEL · WMT En->De newstest2016 | BLEU: 36.9, 36.9, 37.0, 37.0 | No clear effect |
Closest papers by meaning
- Data Diversification: A Simple Strategy For Neural Machine Translation (NeurIPS 2020 · similarity 0.88)
- A Paradigm Shift in Machine Translation: Boosting Translation Performance of Large Language Models (ICLR 2024 · similarity 0.86)
- EBBS: An Ensemble with Bi-Level Beam Search for Zero-Shot Machine Translation (AAAI 2025 · similarity 0.86)
- Learning from others' mistakes: Finetuning machine translation models with span-level error annotations (ICML 2025 · similarity 0.86)
- Ensemble Learning for Heterogeneous Large Language Models with Deep Parallel Collaboration (NeurIPS 2024 · similarity 0.86)
- Calibrating Translation Decoding with Quality Estimation on LLMs (NeurIPS 2025 · 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.