Search papers, labs, and topics across Lattice.
1
0
2
Every configuration of LLM agents hallucinated skill names, with rates reaching over 43% in real-world scenarios, exposing a critical vulnerability that attackers can exploit.
