Last Translation Benchmark: 3,456 Crowdsourced Examples That Break SOTA MT

Researchers release the Last Translation Benchmark: 3,456 crowdsourced hard examples that defeat state-of-the-art MT models, with failures extending beyond figurative language and pointing to a need for better multilingual cultural reasoning.

2026-09-04 ~ 2026-09-04 · 3 related posts

• Machine Translation Is Not Solved: Last Translation Benchmark Released — zouharvi · 2026-09-04
• Last Translation Benchmark: 3456 Crowdsourced Examples Break SOTA Translation Models — zouharvi · 2026-09-04
• Translation Models Fail Beyond Figurative Language, Says Benchmark Author — zouharvi · 2026-09-04