Preprint
Oct 2026
PDScope: Breaking the Prefill-Decode Seesaw for Memory-Bound MoE Inference
PDScope achieves 2.5-10.3x higher prefill throughput and 2.1-6.1x higher decode throughput than existing systems, showing that phase-aware co-design unlocks efficient MoE inference under memory constraints and extends naturally to multi-batch settings.
En-Da Yu, De-Zun Dong, Zhao-Ning Zhang et al.
· 0 citations