Search papers, labs, and topics across Lattice.
Affiliation:
1
0
2
9
Achieving a 3.1x speedup in inference time while retaining 93.8% of performance, PACE revolutionizes how VLMs handle visual token efficiency.
