Search papers, labs, and topics across Lattice.
1
0
3
4
Robot video world models can be significantly improved by distilling a multimodal reward function and stabilizing long-horizon inference, leading to better instruction following and manipulation accuracy.
