Search papers, labs, and topics across Lattice.
Affiliation:
2
0
7
State information in LLM-driven agents can be exploited, turning their task execution capabilities into a potential attack surface.
