Search papers, labs, and topics across Lattice.
2
0
3
2
Visual updates in DeltaV cut token generation by over half while boosting reasoning accuracy, challenging the need for full-image outputs in multimodal models.
Switch-Reasoner reveals that adaptive reasoning selection can enhance MLLM performance by reducing unnecessary cognitive load while maintaining accuracy.
