Search papers, labs, and topics across Lattice.
1
0
3
Grey-box fuzzing of LLM agents, guided by tool invocation sequences, reveals significantly more prompt injection vulnerabilities and malicious behaviors than black-box testing alone.
