Search papers, labs, and topics across Lattice.
1
0
3
13
Robots can now learn better world models through unsupervised self-play, outperforming models trained on human data by 40% in failure prediction and 65% in real-world RL.
