Search papers, labs, and topics across Lattice.
3
0
10
State information in LLM-driven agents can be exploited, turning their task execution capabilities into a potential attack surface.
