I am a first-year PhD student at The Chinese University of Hong Kong, advised by Prof. Weiyang Liu. I received my bachelor's degree with honors from the School of Integrated Circuits and Zhiyuan College at Shanghai Jiao Tong University.
Previously, I was a research intern at Microsoft Research Asia (MSRA).
Breaking the Exactness Barrier: Interleaved DeepSeek Sparse Attention for Efficient Long Context Reasoning
TL;DR: Top-k selection in sparse attention is difficult to optimize in distributed systems because it introduces global dependencies. We interleave the context across GPUs and perform top-k selection locally on each GPU, eliminating the global dependency.
SEDG: Stitch-Compatible End-to-End Layout Decomposition Based on Graph Neural Network
TL;DR: Layout decomposition is an important step in chip manufacturing. We explore a more efficient decomposition method using heterogeneous graph neural networks.