Search papers, labs, and topics across Lattice.
1
0
2
Robots can recover from unexpected mid-episode failures up to 87% faster by adding a small, carefully constrained control signal on top of a pre-trained policy, without any further training.
