In March 2023, Tom Kocmi and Christian Federmann from Microsoft demonstrated that large language models (LLM) can be prompted to assess the quality of machine translation (MT), achieving state-of-the-art (SOTA) performance in assessing system-level quality.