📄 中文摘要
大语言模型（LLMs）在危机准备和响应中的应用日益受到关注，尤其是在多语言沟通方面。然而，它们在高风险危机情境中的适用性尚未得到充分评估。本研究考察了最先进的LLMs和机器翻译系统在危机领域翻译中的表现，重点关注紧迫性这一有效危机沟通和分诊的关键属性。利用多语言危机数据和一个新引入的覆盖32种语言的紧迫性标注数据集，研究表明，无论是专用翻译模型还是LLMs，均表现出显著的性能下降和不稳定性。即使在语言上合格的翻译也可能无法有效传达紧迫性。
基于大语言模型的自动翻译与危机情境中的紧迫性
出处: LLM-Powered Automatic Translation and Urgency in Crisis Scenarios
发布: 2026年2月17日
📄 English Summary
LLM-Powered Automatic Translation and Urgency in Crisis Scenarios
Large language models (LLMs) are increasingly being utilized for crisis preparedness and response, particularly in multilingual communication. However, their effectiveness in high-stakes crisis contexts remains inadequately assessed. This research evaluates the performance of state-of-the-art LLMs and machine translation systems in the domain of crisis translation, emphasizing the preservation of urgency, a critical element for effective crisis communication and triaging. Utilizing multilingual crisis data and a newly introduced urgency-annotated dataset encompassing over 32 languages, findings reveal that both dedicated translation models and LLMs experience significant performance degradation and instability. Notably, even linguistically adequate translations may fail to convey the necessary sense of urgency.