Zhejiang University Team Creates Visual Reasoning System That Lets Robots 'Think With Their Eyes' — 22x Faster Than Text
Researchers from Zhejiang University, in partnership with Cornell University, the National University of Singapore, and Xidian University, have developed a novel visual reasoning system for robots. This system, named VisualThink-VLA, allows robots to process visual information directly, bypassing the need for language-based internal thought processes. The new system demonstrates a significant speed enhancement, being 22.8 times faster than traditional text-based reasoning methods, while also achieving superior accuracy. AI
IMPACT Enables robots to process visual information directly, potentially speeding up complex tasks and improving their interaction with the physical world.