![[2602.21212] Disaster Question Answering with LoRA Efficiency and Accurate End Position](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.21212] Disaster Question Answering with LoRA Efficiency and Accurate End Position
Summary
This paper presents a disaster-focused question answering system optimized for Japanese disaster scenarios, achieving high accuracy with a lightweight model architecture.
Why It Matters
Natural disasters pose significant challenges for effective communication and response. This research addresses the critical need for accurate information dissemination during such events, leveraging advanced machine learning techniques to enhance disaster response capabilities. By focusing on Japanese contexts, it provides insights that could be adapted globally, improving preparedness and response strategies.
Key Takeaways
- Introduces a disaster question answering system tailored for Japanese scenarios.
- Achieves 70.4% End Position accuracy with a lightweight model using LoRA optimization.
- Combines Japanese BERT-base with Bi-LSTM for improved contextual understanding.
- Highlights the importance of continual learning and knowledge updates in disaster response.
- Identifies future challenges in developing efficient AI applications for disaster situations.
Computer Science > Computation and Language arXiv:2602.21212 (cs) [Submitted on 28 Jan 2026] Title:Disaster Question Answering with LoRA Efficiency and Accurate End Position Authors:Takato Yasuno View a PDF of the paper titled Disaster Question Answering with LoRA Efficiency and Accurate End Position, by Takato Yasuno View PDF HTML (experimental) Abstract:Natural disasters such as earthquakes, torrential rainfall, floods, and volcanic eruptions occur with extremely low frequency and affect limited geographic areas. When individuals face disaster situations, they often experience confusion and lack the domain-specific knowledge and experience necessary to determine appropriate responses and actions. While disaster information is continuously updated, even when utilizing RAG search and large language models for inquiries, obtaining relevant domain knowledge about natural disasters and experiences similar to one's specific situation is not guaranteed. When hallucinations are included in disaster question answering, artificial misinformation may spread and exacerbate confusion. This work introduces a disaster-focused question answering system based on Japanese disaster situations and response experiences. Utilizing the cl-tohoku/bert-base-japanese-v3 + Bi-LSTM + Enhanced Position Heads architecture with LoRA efficiency optimization, we achieved 70.4\% End Position accuracy with only 5.7\% of the total parameters (6.7M/117M). Experimental results demonstrate that the combinatio...
