Search papers, labs, and topics across Lattice.
Corresponding author
1
0
3
2
VLMs can be made robust to adversarial image attacks without retraining, using a clever text-augmentation scheme at inference time.
