AGC-Drive: A Large-Scale Dataset for Real-World Aerial-Ground Collaboration in Driving Scenarios

WarNav is a new dataset designed for benchmarking semantic segmentation models for autonomous vehicle navigation in conflict-affected environments. It utilizes images from the DATTALION repository to fill a gap between urban driving datasets and the unique challenges faced in war zones. The dataset addresses methodological challenges and provides baseline results using state-of-the-art models, guiding future research in extreme operational contexts.

This study evaluates the understanding of road safety principles by multi-modal large language models (LLMs), particularly in the context of autonomous vehicles. Using a curated dataset of traffic signs and safety norms from school textbooks, the research reveals that these models struggle with safety reasoning, highlighting significant gaps between human learning and model interpretation. The findings suggest a need for further research to address these performance deficiencies in AI systems governing autonomous vehicles.