Google researchers propose a two-step method to make human translation evaluation more reliable — without doubling costs.

When traveling, AI translators are still more cumbersome than the universal language of pointing fingers and Googling.

In an October 28, 2025, paper, researchers from Alibaba uncovered major reliability issues in multilingual large language models (LLMs) used for AI translation, warning that even top-tier models ...