A company uses generative AI with large language models (LLMs) to translate training manuals from English into other languages. They need to assess the accuracy of the generated translated text. Which model evaluation approach satisfies this need?