1
2
MobileMoE: Scaling On-Device Mixture of Experts
首个将混合专家模型高效部署到移动设备上的架构，实现低延迟与资源友好的AI推理。
arXiv:2605.27358v1 Announce Type: cross Abstract: Mixture-of-Experts (MoE) has become the de facto architecture for hundred-billion-parameter language…