ConceptVTR / Visual Token Routing
Visual Token Routing
A joint training framework for vision-language models that selectively retains key visual tokens through a lightweight spatiotemporal router to reduce the cost of long-context inference.
Timeline (last 90 days)
Nov 5
The researchers propose Visual Token Routing (VTR), a joint training framework for learning sparse visual representations that retain only the information needed for action prediction.
Unverified50%