CityLens: Benchmarking Large Language-Vision Models for Urban Socioeconomic Sensing

The study introduces CityLens, a benchmark dataset designed to evaluate large language-vision models' ability to predict urban socioeconomic indicators from satellite and street view images across 17 cities worldwide. While LLVMs show promising perceptual and reasoning skills, they still face limitations in accurately predicting complex urban socioeconomic factors, highlighting areas for future improvement.