Microsoft Research Blog
AsgardBench: A benchmark for visually grounded interactive planning
Imagine a robot tasked with cleaning a kitchen. It needs to observe its environment, decide what to do, and adjust when things don’t go as expected, for example, when the mug it was tasked to…
Microsoft Research Blog
GroundedPlanBench: Spatially grounded long-horizon task planning for robot manipulation
Vision-language models (VLMs) use images and text to plan robot actions, but they still struggle to decide what actions to take and where to take them. Most systems split these decisions into two steps: a…