Alibaba International Unveils Groundbreaking AI Translation Evaluation List: TransBench
2025-05-22 / Read about 0 minute
Author:小编   

In a collaborative effort with multiple partners, Alibaba International has introduced the AI translation evaluation list, named TransBench, along with the OpenCompass platform. TransBench enhances traditional evaluation metrics by incorporating novel indicators such as hallucination rate and culturally taboo words. For the first time, it specifically tailors evaluation data and methodologies to distinct segments of the translation industry. All indicators are grounded in real-world scenario feedback, and the evaluation methods have been made openly available through open-source initiatives. This groundbreaking move is poised to accelerate the advancement of AI translation technology.