Search papers, labs, and topics across Lattice.
1
0
3
2
Visual token dominance is the hidden culprit behind LVLM inference inefficiency, and this paper dissects the problem to reveal how to navigate the fidelity-efficiency tradeoff.
