Search papers, labs, and topics across Lattice.
1
0
3
35
Training a single model across text, images, video, 3D geometry, and hidden representations unlocks "Context Unrolling," where the model reasons across modalities to improve reasoning fidelity.
