Search papers, labs, and topics across Lattice.
1
0
4
Genomic language models leak substantial training data through repeated generations, with a black-box attack recovering over 500 bp segments in nearly 10% of trials.