Overview
This project studies how translation errors influence multilingual LLM evaluation. It connects translation quality estimation with benchmark validity and model comparison.
Associated publication
Quantifying the Impact of Translation Errors on Multilingual LLM Evaluation.
Under review / ACL presentation planned.
Resources
| Resource | Link |
|---|---|
| Paper | TODO |
| Code | TODO |
| Dataset | TODO |
| Demo | TODO |
| Blog post | ACL 2026 preview |
Research questions
- Which translation errors have the strongest impact on benchmark outcomes?
- Can translation quality estimation help flag unreliable benchmark instances?
- How should multilingual evaluation reports communicate translation-related uncertainty?