Search papers, labs, and topics across Lattice.
3
0
5
8
Skip the costly robot teleoperation data: ZeroWBC learns surprisingly natural humanoid control policies directly from human egocentric videos.
Diffusion policies get a 19% adaptability boost in dynamic robotic tasks, without retraining, thanks to a closed-loop architecture that dynamically corrects actions in real-time.
VLMs can be taught to understand the physics of image degradation well enough to control diffusion models for zero-shot image restoration, without fine-tuning the generative backbone.
