Search papers, labs, and topics across Lattice.
1
0
2
Even state-of-the-art machine unlearning algorithms can leak sensitive information about forgotten categories through model parameters and inversion, enabling surprisingly effective attacks.