Preprint
Sep 2026
BigMoMo: Efficient Inference of Large-Scale MoE with Speculative Decoding on Mobile Devices
A mobile MoE runtime that exploits the multi-token verification window of speculative decoding to decouple expert movement from single-token execution, enabling weight reuse, contiguous flash reads, and load-compute overlap.
Mao-Liang Li, H. Zou, Tao Han et al.
· 0 citations