06/08/2026
Two of our most-read pieces in GALA's Resource Center ask the same question: How do you actually measure AI translation quality? A third asks whether the LLMs doing the measuring can be trusted to do it.
𝐇𝐨𝐰 𝐭𝐨 𝐓𝐞𝐬𝐭 𝐚𝐧𝐝 𝐒𝐜𝐨𝐫𝐞 𝐀𝐈 𝐓𝐫𝐚𝐧𝐬𝐥𝐚𝐭𝐢𝐨𝐧 𝐐𝐮𝐚𝐥𝐢𝐭𝐲 (contributed by Margarita Shvetsova) walks through a full testing workflow: sample text, linguist review, post-edited reference translation, then automated metrics run against it.
Read it here: https://ow.ly/Yawl50ZkYH9
𝐍𝐞𝐰 𝐌𝐞𝐭𝐫𝐢𝐜𝐬 𝐟𝐨𝐫 𝐀𝐈 𝐓𝐫𝐚𝐧𝐬𝐥𝐚𝐭𝐢𝐨𝐧 𝐐𝐮𝐚𝐥𝐢𝐭𝐲 (contributed by Nadine Lai) takes a different angle: Which metric should you reach for and when? Budget and content type decide the answer.
Read it here: https://ow.ly/xUn650ZkYHa
𝐓𝐫𝐚𝐧𝐬𝐥𝐚𝐭𝐢𝐨𝐧 𝐐𝐮𝐚𝐥𝐢𝐭𝐲 𝐀𝐈 𝐢𝐧 𝐭𝐡𝐞 𝐋𝐋𝐌 𝐄𝐫𝐚 moves the question one step further. Alon Lavie's recorded presentation covers the 2025 WMT Shared Task on Automated Translation Quality Evaluation, testing how LLM-based evaluators perform at judging translation output themselves.
Watch the video here: https://ow.ly/HtyN50ZkYHb