Search papers, labs, and topics across Lattice.
1
0
3
Frontier multimodal models that master complex single-image visual reasoning collapse to sub-35% accuracy when asked to detect basic low-level noise and texture differences between two images.
