Search papers, labs, and topics across Lattice.
Carnegie Mellon University
2
0
3
Alignment-based defenses can effectively shield LLMs from property inference attacks without requiring access to original training data.
