Search papers, labs, and topics across Lattice.
1
0
2
6
Adversarial text can increase localization errors in MLLMs by nearly five times, revealing critical vulnerabilities in multimodal reasoning.
