BROWSE
Detail
Conference Paper
Evaluation of LLM-Based VQA Models for Enhancing Battlefield Situation Awareness
Cited 0 time in 
Share 
Share
- Issue Date
- 2025-02
- Citation
- International Conference on Big Data and Smart Computing (BigComp) 2025, pp.421-423
- Language
- English
- Type
- Conference Paper
- Abstract
- Multimodal Visual Question Answering (VQA) has emerged as a critical technology for integrating image data and text-based queries to enhance situational awareness in complex environments. By leveraging the capabilities of Large Language Models (LLMs), VQA systems can process external knowledge and provide accurate, context-aware responses. This study explores the development of Korean-supporting multimodal VQA models, focusing on the architectural integration of vision encoders, projection layers, and LLMs. Through pre-training and fine-tuning, the research aims to evaluate these models for tasks requiring language adaptation and external knowledge handling in military scenarios.
- KSP Keywords
- Architectural integration, Battlefield situation, Context aware, External knowledge, Fine-tuning, Image data, Pre-Training, Situation awareness(SA), Situational Awareness, Visual Question Answering, complex environment