Languages Still Left Behind: Toward a Better Multilingual Machine Translation Benchmark

ACL ARR 2025 May Submission7735 Authors

Abstract:

Multilingual machine translation (MT) benchmarks play a central role in evaluating the capabilities of modern MT systems. Among them, the FLORES+ benchmark is widely used, offering English-to-many translation data for over 200 languages, curated with strict quality control protocols. However, we study data in four languages (Asante Twi, Japanese, Jinghpaw, and South Azerbaijani) and uncover critical shortcomings in the benchmark’s suitability for truly multilingual evaluation. Human assessments reveal that many translations fall below the claimed 90% quality standard, and the annotators report that source sentences are often too domain-specific and culturally biased toward the English-speaking world. We further demonstrate that simple heuristics, such as copying named entities, can yield non-trivial
BLEU scores, suggesting vulnerabilities in the evaluation protocol. Notably, we show that MT models trained on naturalistic data perform poorly on FLORES+ while achieving significant gains on our domain-relevant evaluation set. Based on these findings, we advocate for multilingual MT benchmarks that use domain-general, named-entity-agnostic, and culturally neutral source texts to better reflect real-world translation challenges.

Paper Type: Long

Research Area: Machine Translation

Research Area Keywords: human evaluation, multilingual MT, multilingual benchmarks, multilingual evaluation, less-resourced languages

Contribution Types: Reproduction study, Approaches to low-resource settings, Data resources, Data analysis, Position papers

Languages Studied: Asante Twi, English, Japanese, Jinghpaw, South Azerbaijani

Submission Number: 7735