Search papers, labs, and topics across Lattice.
1
0
3
LLMs are hiding interpretable semantic structures in plain sight: just five lines of code and an SVD can expose their training data secrets and even ethically problematic content.