Search papers, labs, and topics across Lattice.
1
0
3
Achieving a 90% boost in prefill throughput for MoE models could redefine the efficiency of large-scale language model serving.
