Preprint
Aug 2026
GST-Bench: Can VLMs Develop Global Spatial Awareness from Video?
The Global-Spatial-Temporal Benchmark (GST-Bench), a VQA benchmark for global spatial intelligence in video understanding, comprising human-verified questions derived from 6,790 minutes of synthetically generated video, is introduced and it is found that models, despite strong local spatial understanding under the same...
Qifeng Zhang, Kai-Xiang Huang, Heng Dong et al.
· 1 citation