Search papers, labs, and topics across Lattice.
1
0
3
Ditching pixel-space translation unlocks a unified model (LatentUM) that reasons across modalities with SOTA results, opening doors to more efficient and aligned visual AI.
