Search papers, labs, and topics across Lattice.
1
0
3
CAEE slashes inference latency in MoE models by intelligently pruning low-value experts, achieving efficiency gains without sacrificing accuracy.
