Lost in the Source Language: How Large Language Models Evaluate the Quality of Machine Translation
This addresses the problem of understanding LLM evaluation mechanisms for researchers in NLP, but it is incremental as it builds on existing knowledge without major breakthroughs.
This study investigated how Large Language Models (LLMs) use source and reference data in machine translation evaluation, finding that reference information improves accuracy while source information can be counterproductive, indicating LLMs' limited cross-lingual capability in this task.
This study investigates how Large Language Models (LLMs) leverage source and reference data in machine translation evaluation task, aiming to better understand the mechanisms behind their remarkable performance in this task. We design the controlled experiments across various input modes and model types, and employ both coarse-grained and fine-grained prompts to discern the utility of source versus reference information. We find that reference information significantly enhances the evaluation accuracy, while surprisingly, source information sometimes is counterproductive, indicating LLMs' inability to fully leverage the cross-lingual capability when evaluating translations. Further analysis of the fine-grained evaluation and fine-tuning experiments show similar results. These findings also suggest a potential research direction for LLMs that fully exploits the cross-lingual capability of LLMs to achieve better performance in machine translation evaluation tasks.