Search papers, labs, and topics across Lattice.
1
0
3
By framing adversarial training as a zero-sum Markov game, ADV-0 finds more diverse safety-critical failures in autonomous driving systems, leading to significantly improved generalization against unseen long-tail risks.
