Towards Explainable & Efficient Automated Crisis Response
Loading...
Date
2025-05
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
The Ohio State University
Abstract
Recent advancements in Natural Language Processing and Computer Vision have been largely propelled by the advent of foundational models, such as large language models (LLMs) and state-of-the-art vision systems. These models have achieved unprecedented performance in a variety of tasks, ranging from translation and summarization to image segmentation and classification. In the context of crisis response, where multi-modal data streams are abundant, the utilization of these models offers a promising way to assimilate trustworthy and contextually grounded information. Nevertheless, the inherent black-box nature of these models poses significant challenges in terms of explainability, and their substantial computational demands restrict deployment in resource-constrained environments, particularly where dedicated GPU hardware is unavailable.
In this study, we investigate methodologies aimed at mitigating these critical limitations. Firstly, we propose a framework that leverages LLMs to facilitate decision-making processes underpinned by high-quality factual data. Secondly, we present a series of optimizations applied to a large-scale vision model to enhance resource efficiency, and discuss its applicability in Knowledge Gap Identification.
Our results indicate that addressing both the explainability and efficiency deficits inherent in foundational models improves the utility of these capable models in real-world crisis response scenarios.
Description
Keywords
Computer Vision, Natural Language Processing, AI, Machine Learning, Crisis Response, Disaster Response