Firehose

This paper investigates whether visual foundation models can support spatial state representations that are updated under self-motion and used to guide action, which is crucial for tasks like spatial perception and object tracking. Practitioners working on visual perception, object tracking, or action planning might care about this research.