Search papers, labs, and topics across Lattice.
Xi鈥檃n Jiaotong University
1
0
3
2
MMLMs can be made 99% safer against harmful multimodal inputs without sacrificing utility, thanks to a novel calibration approach.
