Search papers, labs, and topics across Lattice.
Corresponding author are Bo Cheng and Soujanya Poria
2
0
4
5
By explicitly grounding reasoning steps to visual objects, Chain-of-Glimpse enables more accurate and interpretable video understanding, outperforming object-agnostic methods on multiple benchmarks.
Imagine creating high-fidelity, navigable 3D worlds from just a text prompt or a single image – HY-World 2.0 makes it a reality.
