Search papers, labs, and topics across Lattice.
1
0
2
Current agents are alarmingly susceptible to skill-based attacks, with success rates reaching over 86%, exposing a critical vulnerability in AI safety.
