Search papers, labs, and topics across Lattice.
1
0
2
13
Mage-VL slashes visual token usage by over 75% while enhancing real-time multimodal performance, outperforming larger models in video understanding.
