We propose a benchmark to evaluate LLMs' spatiotemporal reasoning capabilities

We track objects through transformations while detecting and describing these state changes.