Congratulations! Sunhong’s paper “CINELL: An Energy-Efficient Compute-In/Near-Memory eDRAM Processor for Sparse Transformer-Based Large Language Models” is highlighted by IEEE Transactions on VLSI Systems (TVLSI)!

This paper introduces CINELL, a compute-in/near-memory eDRAM processor designed to tackle the heavy computation and memory bandwidth demands of transformer-based LLMs. With attention block fusion, a CINM architecture, and compute-in-memory acceleration, CINELL delivers major gains in latency, memory access, and energy efficiency — enabling practical, high-performance LLM inference.