arxivcs.CV2026-06-26
AirGroundBench: Probing Spatial Intelligence in Multimodal Large Models under Heterogeneous Multi-View Embodied Collaboration
Haotian Li, Yida Wang, Leyuan Wang, Jinshan Lai, Keyang Wang, Zonghao Guo, et al.
In recent years, multimodal large language models (MLLMs) have shown strong potential for embodied intelligence, yet their ability to maintain geometrically consistent spatial understanding across heterogeneous views remains under-evaluated. Existing benchmarks largely focus on s…