Search papers, labs, and topics across Lattice.
1
0
3
2
MLLMs struggle to ground cultural values in visual scenes, losing ~7% accuracy compared to text-only inputs, even when they understand the visual content.
