HyMoE-CIM: A Hybrid-Bit SRAM/ReRAM CIM Architecture with Hardware-Aware Token Dispatch for Mixture-of-Experts LLM Inference
ICCD 2026IEEE International Conference on Computer Design (ICCD)
This paper presents a hybrid-bit SRAM/ReRAM CIM architecture with hardware-aware token dispatch for MoE LLM inference.
Citation
Zhibiao Xue, Xu Chen, Kunming Shao, and collaborators. HyMoE-CIM: A Hybrid-Bit SRAM/ReRAM CIM Architecture with Hardware-Aware Token Dispatch for Mixture-of-Experts LLM Inference. In 2026 IEEE International Conference on Computer Design (ICCD), 2026.