Benchmark RadarUrbanWell
UrbanWell provides a dataset and evaluation protocol for assessing spatio-temporal reasoning in multimodal large language models using satellite and street view imagery across 38 cities, covering environmental, accessibility, urban form, vitality, and subjective perception indicators with tasks in static prediction, forecasting, and trend classification.
- Released
- 2026-06-14
- Readiness
- Runnable
- Primary field
- General AI
Why it matters
UrbanWell addresses the lack of standardized benchmarks for multimodal urban wellbeing analytics, enabling consistent comparison of MLLMs on tasks requiring joint spatial and temporal understanding. This supports progress in urban intelligence applications such as planning and policy evaluation.
Motivation
Understanding urban wellbeing from multimodal data requires integrating heterogeneous spatial and temporal signals, posing significant challenges for current multimodal large language models (MLLMs).
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.