김상엽Sangyeob Kim

DPA-MoE: A Runtime Precision Control Framework for MoE Inference

J. Kim, H. Lee, S. Shin, H. Kim, D. Lee, Sangyeob Kim

MonarchGRU: A Low-Footprint Recurrent Neural Network for Audio Keyword Spotting System

S. Bhowmick, Sangyeob Kim, Q. Chen, C. Gao, K. Kim

NeuLPU: A 3.6 mW Spiking LLM Processing Unit for On-device Long-context Inference

S. Lee, S. Ha, S. Kim, W. Jo, Sangyeob Kim, Hoi-jun yoo

A 4.69mW LLM Processor with Binary/Ternary Weights for Billion-Parameter Llama Model

Sangyeob Kim, J. Lee, B. Kim, Hoi-Jun Yoo

A 65.1 TOPS/W Digital CIM Processor for Ultra-Low-Bit Transformers with Multiplexer-Based Adder and Scaling Factor-Based Reordering

N.Lee, Sangyeob Kim, S.Hong, J.Choi, Hoi-Jun Yoo

BROCA: A 52.4-559.2mW Mobile Social Agent System-on-Chip with Adaptive-Bit-Truncate Unit and Acoustic-Cluster Bit-Grouping

W. Jo, S. Hong, J. Choi, B. Kwon, H. Sang, D. Im, Sangyeob Kim, S. Kim, Hoi-Jun Yoo

BROCA: A Low-power and Low-latency Conversational Agent RISC-V System-on-Chip for Voice-interactive Mobile Devices

W. Jo, S. Hong, J. Choi, B. Kwon, H. Sang, D. Im, Sangyeob Kim, S. Kim, C. Jeong, Y. Moon, Hoi-Jun Yo

C-Transformer: A Homogeneous DNN-Transformer/Spiking-Transformer Processor for Large-Language-Models

Sangyeob Kim, S.Kim, W.Jo, S.Kim, S.Hong, N.Lee, J.Lee, Hoi-Jun Yoo