【摘要】 大语言模型驱动的翻译智能体因实现方式不同，可归纳为：基于提示工程的翻译智能体、基于检索增强生成的翻译智能体和基于计算机辅助翻译的翻译智能体。文章在综述大语言模型、智能体和翻译智能体的基础上，尝试阐述三类翻译智能体的工作原理和技术实现，以法律文本为例，评估智能体的译文质量和实践效果。结果表明，大语言模型驱动的翻译智能体可提升翻译质量，提高翻译效率，应对复杂任务。该研究为翻译实践和教学提供了新视角，为翻译技术创新发展和翻译产业转型升级提供了思路和方法。更多还原

【Abstract】 This study categorized LLM-powered translation agents( Trans Agents) into three groups based on their implementation approaches: Prompt Engineering-based Trans Agents, Retrieval Augmented Generation-based Trans Agents,and Computer-Assisted Translation-based Trans Agents. Upon a comprehensive review of the literature on LLMs,agents,and Trans Agents,this paper elucidated the operational principles and technical implementations of these Trans Agents. Legal texts were then selected as specific cases to evaluate and compare their translation performance. The findings demonstrated that LLM-based Trans Agents hold significant promise in improving translation accuracy,enhancing translation efficiency,and coping with complex translation tasks. The research introduced a fresh perspective to translation practice and education,along with novel ideas and approaches for advancing translation technology and upgrading translation industry.更多还原