We presented an Accelerator-in-Memory (AiM) device and AiM-based LLM inference acceleration system. LLM inference can be divided into prompt phase and response phase. Considering the characteristics ...