Search papers, labs, and topics across Lattice.
Adelaide University
1
0
3
Corrupting the imagination in VLA policies is 60 times more effective than random perturbations, exposing a critical vulnerability in safety mechanisms.
