AeroGround: A Comprehensive Benchmark for Aerial-Ground Collaborative Reasoning

A new benchmark, AeroGround, evaluates vision-language models (VLMs) on aerial-ground collaborative reasoning tasks, revealing a significant performance gap: the best model achieves 54.4% average accuracy versus 93.3% for humans. The benchmark, built on a simulated dataset with approximately 29,000 multimodal observation groups and 2,250 question-answering instances, tests cross-view correspondence, spatial understanding, and reasoning across 16 pretrained VLMs and two domain-adapted variants. The findings highlight current model limitations and aim to guide development of more capable aerial-ground collaborative embodied intelligence systems.

arXiv:2608.14721v1 Announce Type: new Abstract: Vision-language models VLMs have been widely employed in understanding and reasoning tasks for unmanned aerial vehicles UAVs . Existing UAV benchmarks primarily focus on aerial-view scenarios. However, whether current VLMs can perform well on understanding and reasoning tasks in aerial-ground collaborative scenarios which are practical in real-world applications like rescue and infrastructure inspection remains underexplored. To address this gap, we introduce AeroGround, a comprehensive benchmark for evaluating VLMs in aerial-ground collaborative reasoning. AeroGround is built upon a simulated aerial-ground dataset containing approximately 29,000 multimodal observation groups from diverse open environments, and provides 2,250 high-quality question-answering instances covering cross-view correspondence, spatial understanding, and reasoning. Experiments on 16 pretrained VLMs, together with two domain-adapted variants, reveal a substantial gap between current models and human performance: the best model achieves an average accuracy of 54.4%, whereas humans reach 93.3%. By systematically revealing the strengths and limitations of existing models in aerial-ground collaborative reasoning, AeroGround provides a foundation for developing more capable aerial-ground collaborative embodied intelligence systems.