CityCube: Benchmarking Cross-view Spatial Reasoning on Vision-Language Models in Urban Environments
This addresses the problem of evaluating spatial reasoning in AI for urban applications, but it is incremental as it builds on existing benchmark work by focusing on a specific domain.
The authors tackled the lack of benchmarks for cross-view spatial reasoning in urban environments by introducing CityCube, a dataset with 5,022 annotated multi-view QA pairs, and found that large-scale VLMs achieve only 54.1% accuracy, 34.2% below human performance, while fine-tuned small-scale models exceed 60.0%.
Cross-view spatial reasoning is essential for embodied AI, underpinning spatial understanding, mental simulation and planning in complex environments. Existing benchmarks primarily emphasize indoor or street settings, overlooking the unique challenges of open-ended urban spaces characterized by rich semantics, complex geometries, and view variations. To address this, we introduce CityCube, a systematic benchmark designed to probe cross-view reasoning capabilities of current VLMs in urban settings. CityCube integrates four viewpoint dynamics to mimic camera movements and spans a wide spectrum of perspectives from multiple platforms, e.g., vehicles, drones and satellites. For a comprehensive assessment, it features 5,022 meticulously annotated multi-view QA pairs categorized into five cognitive dimensions and three spatial relation expressions. A comprehensive evaluation of 33 VLMs reveals a significant performance disparity with humans: even large-scale models struggle to exceed 54.1% accuracy, remaining 34.2% below human performance. By contrast, small-scale fine-tuned VLMs achieve over 60.0% accuracy, highlighting the necessity of our benchmark. Further analyses indicate the task correlations and fundamental cognitive disparity between VLMs and human-like reasoning.