📄 中文摘要
自然灾害如地震、暴雨、洪水和火山喷发发生频率极低，且影响的地理区域有限。在灾难情况下，个体常常感到困惑，缺乏必要的领域特定知识和经验来判断适当的应对措施。尽管灾难信息不断更新，利用RAG搜索和大型语言模型进行查询时，获取与特定情况相关的领域知识并不总是能得到保证。灾难问答中如果出现幻觉，可能导致人工错误信息的传播，从而加剧混乱。该研究提出了一种新的方法，旨在提高灾难问答的准确性和效率，特别是在处理低频事件时，确保提供可靠的信息和建议。
基于LoRA效率和准确终点的灾难问答
📄 English Summary
Disaster Question Answering with LoRA Efficiency and Accurate End Position
Natural disasters such as earthquakes, heavy rainfall, floods, and volcanic eruptions occur with extremely low frequency and affect limited geographic areas. Individuals often experience confusion during disaster situations and lack the domain-specific knowledge and experience necessary to determine appropriate responses. Although disaster information is continuously updated, utilizing RAG search and large language models does not guarantee access to relevant domain knowledge tailored to specific situations. The presence of hallucinations in disaster question answering can lead to the spread of artificial misinformation, exacerbating confusion. This research introduces a novel approach aimed at enhancing the accuracy and efficiency of disaster question answering, particularly in the context of low-frequency events, ensuring the provision of reliable information and guidance.