Search papers, labs, and topics across Lattice.
Texas A&M University
2
0
2
Privacy-sensitive visual and textual information can be recovered from deep-layer LVLM hidden states, exposing significant risks in collaborative inference.
Backdoor attacks can become virtually immune to pruning and significantly more resilient to fine-tuning by carefully crafting triggers that push poisoned samples into low-density regions of the clean data distribution.
