TEaR: Improving LLM-based Machine Translation with Systematic Self-Refinement
Zhaopeng Feng 1 Yan Zhang 2∗ Hao Li22{}^{~{}2}start_FLOATSUPERSCRIPT 2 end_FLOATSUPERSCRIPT Bei Wu22{}^{~{}2}start_FLOATSUPERSCRIPT 2 end_FLOATSUPERSCRIPT Jiayu Liao22{}^{~{}2}start_FLOATSUPERSCRIPT 2 end_FLOATSUPERSCRIPT Wenqiang Liu22{}^{~{}2}start_FLOATSUPERSCRIPT 2 end_FLOATSUPERSCRIPT Jun Lang22{}^{~{}2}start_FLOATSUPERSCRIPT 2 end_FLOATSUPERSCRIPT Yang Feng33{}^{~{}3}start_FLOATSUPERSCRIPT 3 end_FLOATSUPERSCRIPT Jian Wu11{}^{~{}1}start_FLOATSUPERSCRIPT 1 end_FLOATSUPERSCRIPT Zuozhu Liu 1 1Zhejiang University 2Tencent 3Angelalign Technology Inc. Equally Contributed.Corresponding author.
Abstract
Large Language Models (LLMs) have achieved impressive results in Machine Translation (MT). However, careful evaluations by human reveal that the translations produced by LLMs still contain multiple errors. Importantly, feeding back such error information into the LLMs can lead to self-refinement and result in improved translation performance. Motivated by these insights, we introduce a systematic LLM-based self-refinement translation framework, named TEaR, which stands for Translate, Estimate, and Refine, marking a significant step forward in this direction. Our findings demonstrate that 1) our self-refinement framework successfully assists LLMs in improving their translation quality across a wide range of languages, whether it’s from high-resource languages to low-resource ones or whether it’s English-centric or centered around other languages; 2) TEaR exhibits superior systematicity and interpretability; 3) different estimation strategies yield varied impacts, directly affecting the effectiveness of the final corrections. Additionally, traditional neural translation models and evaluation models operate separately, often focusing on singular tasks due to their limited capabilities
中文速览
大型语言模型（LLM）虽然在机器翻译（Machine Translation, MT）上表现亮眼，但人工检查发现其译文仍存在不少错误，而把这些错误反馈给模型后它能够自我修正、译文质量明显提升。受此启发，研究团队提出了名为 TEaR 的系统性自我精炼翻译框架，分三步走：先用 LLM 生成初译（Translate），再让同一模型按照类人工评估标准对初译进行质量估计并标出错误（Estimate），最后根据估计反馈对译文进行修正（Refine）。在覆盖 17 个翻译方向、高低资源语言及英语与非英语语对的实验中，TEaR 在 COMET、BLEU 等多项自动指标和人工偏好测试上均显著优于直接翻译及现有后编辑基线方法，平均 COMET 提升约 2.48 分。这项工作的价值在于，它为 LLM 的翻译自我修正提供了一套可解释、可复现的系统方案，同时揭示了估计模块质量是制约自我修正效果的关键瓶颈，为后续研究指明了方向。
原文 arXiv:2402.16379;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2402.16379v3