Search papers, labs, and topics across Lattice.
Fudan University ♣ University of California
1
0
3
Visual reasoning models can slash token usage by up to 90% without sacrificing accuracy by adaptively choosing reasoning paths based on task complexity.
