Search papers, labs, and topics across Lattice.
University of Melbourne, Johns Hopkins University
1
0
3
Adversarial attacks on speech models leave tell-tale geometric fingerprints in early representation layers that can be detected without transcripts.
