CECAR: Cache & Expert Co-Aware Routing Accelerates On-Device Inference of MoE LLMswith Hoonki Lee, Byungjun Kim, Inyul Ra, Yongjin Kim +21 versionW_89rfuzt7v1 · current