SpatiaLab is a new test that checks if vision-language models (VLMs) can understand real-world spatial puzzles, like what’s in front, behind, bigger, or reachable.
Long texts are expensive for AI to read because each extra token costs a lot of compute and memory.