Machine Translation Is Not Solved: Last Translation Benchmark Released

zouharvi · x · 2026-09-04

Vilém Zouhar and collaborators argue machine translation is far from solved and release the "Last Translation Benchmark" paper on arXiv. The project uses massive crowdsourcing to collect hard-to-translate examples that break state-of-the-art translation models, enabling more reliable evaluation.

Related event: Last Translation Benchmark: 3,456 Crowdsourced Examples That Break SOTA MT(3 posts)→

Original post →

More from Research

Research channel →