Search papers, labs, and topics across Lattice.
University of Electronic Science and Technology
1
0
2
ST$^2$U reduces restricted-knowledge re-entry by up to 45% while preserving model capabilities, reshaping how we approach test-time unlearning in language models.
