Search papers, labs, and topics across Lattice.
Affiliation:
3
0
6
6
State information in LLM-driven agents can be exploited, turning their task execution capabilities into a potential attack surface.
MMLMs can be made 99% safer against harmful multimodal inputs without sacrificing utility, thanks to a novel calibration approach.
AgentSentry stops indirect prompt injection attacks in LLM agents by pinpointing when the attack takes hold using causality, then surgically removing the malicious influence.
