Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Cross-Lingual Pre-Training Based Transfer for Zero-Shot 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 |
|---|---|---|---|
| Use cross-lingual pre-training with MLM+BRLM-SA instead of pivoting | Zero-shot NMT transfer · Europarl Fr→Es | BLEU: 32.25 -> 37.02 | Helped |
| Use cross-lingual pre-training with MLM+BRLM-SA instead of pivoting | Zero-shot NMT transfer · Europarl De→Fr | BLEU: 27.79 -> 30.66 | Helped |
| Use cross-lingual pre-training with MLM+BRLM-SA instead of pivoting | Zero-shot NMT transfer · Europarl Ro→De | BLEU: 14.74 -> 15.62 | Helped |
| Add back translation to MLM+BRLM-SA | MultiUN zero-shot NMT transfer · MultiUN A-ZST | BLEU: 30.68 -> 34.95 | Helped |
| Add source-pivot bilingual pre-training with TLM (ablation) | MLM zero-shot transfer · Europarl Fr→Es | BLEU: 35.96 -> 36.78 | Helped |
| Add BRLM-SA instead of TLM (ablation) | MLM zero-shot transfer · Europarl Fr→Es | BLEU: 36.78 -> 37.02 | Helped |
| Use soft alignment instead of hard alignment in BRLM (ablation) | MLM+BRLM zero-shot transfer · Europarl Fr→Es | BLEU: 36.30 -> 37.02 | Helped |
| Add source-pivot bilingual pre-training with TLM (ablation) | MLM zero-shot transfer · MultiUN A-ZST | BLEU: 18.36 -> 24.14 | Helped |
| Add BRLM-SA instead of TLM (ablation) | MLM zero-shot transfer · MultiUN A-ZST | BLEU: 24.14 -> 30.68 | Helped |
| Use soft alignment instead of hard alignment in BRLM (ablation) | MLM+BRLM zero-shot transfer · MultiUN A-ZST | BLEU: 25.25 -> 30.68 | Helped |
| Add back translation to MLM+BRLM-SA (ablation) | MultiUN zero-shot transfer · MultiUN Es→Ru | BLEU: 30.47 -> 32.53 | Helped |
| Freeze 4 encoder layers during transfer training (ablation) | MLM+BRLM-SA · MultiUN Ar→Ru | BLEU: 16.09 -> 25.49 | Helped |
Closest papers by meaning
- Evaluating the Cross-Lingual Effectiveness of Massively Multilingual Neural Machine Translation (AAAI 2020 · similarity 0.91)
- Dual Transfer Learning for Neural Machine Translation with Marginal Distribution Regularization (AAAI 2018 · similarity 0.89)
- Cross-Lingual Natural Language Generation via Pre-Training (AAAI 2020 · similarity 0.88)
- Constrained Decoding for Cross-lingual Label Projection (ICLR 2024 · similarity 0.88)
- Zero-Shot Tokenizer Transfer (NeurIPS 2024 · similarity 0.86)
- EBBS: An Ensemble with Bi-Level Beam Search for Zero-Shot Machine Translation (AAAI 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.