Search papers, labs, and topics across Lattice.
Beijing Institute of Technology
1
0
2
Achieving over 90% token reduction in 3D VLMs without sacrificing performance could revolutionize the efficiency of 3D scene understanding.
