Search papers, labs, and topics across Lattice.
1
0
3
3
Weaker autonomous web agents readily trust tampered website content, producing unsafe outputs, while stronger models exhibit better anomaly detection and safer fallback strategies under MITM attacks.
