Search

MindTopo tests the gap between recognising a scene and planning actions

Introduced by Microsoft Research on August 12, the benchmark examines whether vision-language models can preserve spatial relationships while acting.