Research map / Label-efficient and robust learning / Recognition and long-tail
Variance-Aware Machine Translation Test Sets
Runzhe Zhan, Xuebo Liu, Derek F. Wong, Lidia S. Chao
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Filter out low-variance instances using metric-based score variance | WMT test sets · WMT16–WMT20 | system-level correlation with human judgments | Mixed |
| Use VAT with BLEU rather than the original test set | BLEU evaluation · WMT16–WMT20 | |r| | Mixed |
| Use VAT with BERTS-F rather than the original test set | BERTS-F evaluation · WMT16–WMT20 | |r| | Mixed |
| Filter 60% of test instances (ablation) | Variance-aware filtering · WMT20 | Pearson correlation | Improved |
| Filter at an extreme percentage instead of a moderate percentage (ablation) | Variance-aware filtering · WMT20 | Pearson correlation | Hurt |
| Use BERTS-R scores to estimate instance variance for filtering (ablation) | Variance-aware filtering · WMT20 | correlation with human judgments | Improved |
| Create VAT using a random half of submitted systems (ablation) | WMT20 Zh-En and En-Zh · WMT20 Zh-En and En-Zh | Pearson correlation: 0.957 -> 0.961; 0.888 -> 0.893 | Improved |
Closest papers by meaning
- MT-Ranker: Reference-free machine translation evaluation by inter-system ranking (ICLR 2024 · similarity 0.84)
- From Jack of All Trades to Master of One: Specializing LLM-based Autoraters to a Test Set (ICML 2025 · similarity 0.84)
- Contrastive Preference Optimization: Pushing the Boundaries of LLM Performance in Machine Translation (ICML 2024 · similarity 0.83)
- In-Place Test-Time Training (ICLR 2026 · similarity 0.83)
- QUEST: Quality-Aware Metropolis-Hastings Sampling for Machine Translation (NeurIPS 2024 · similarity 0.83)
- A Paradigm Shift in Machine Translation: Boosting Translation Performance of Large Language Models (ICLR 2024 · similarity 0.83)
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.
