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