DishwasherBench requires models to perform three-dimensional spatial reasoning through visual input and common sense, involving constraints such as pot placement, bowl orientation, cup tilt angle, cutlery basket zoning, and object occlusion.