Search papers, labs, and topics across Lattice.
LimX Dynamics Technology Co, Zhejiang University
3
0
6
Forget painstakingly engineering reward functions for every new robotic setup – this method learns a single, reusable reward representation from just five demonstrations that generalizes zero-shot to variations in object, viewpoint, and position.
On-policy reinforcement learning with diffusion policies just got a whole lot easier thanks to a clever method that sidesteps expensive log-likelihood computations.
Achieve 10% higher success rates in robotic manipulation tasks while speeding up inference by 1.5-1.8x by intelligently pruning visual tokens in multi-view Vision-Language-Action models.
