Search papers, labs, and topics across Lattice.
Nanjing University of Aeronautics and Astronautics
1
0
1
A single poisoned skill can stealthily compromise LLM agents, activating only under specific conditions, raising alarms about the security of AI skill supply chains.
