BoxBot Splits Box Opening Between Visual Servoing and a Small VLA

Loading video
Loading videoViam BoxBot opens taped boxes by splitting the stack: image-based visual servoing cuts 1-2 mm tape seams, while a SmolVLA fine-tuned on 125 VR teleop demos peels the deformable flaps, running locally on a Jetson Orin Nano.
Category: arm
Author: @heetezition
Date: 2026-09-23T00:00:00
Duration: 227.62s





