Hi! I am Yikang Yue, a computer science Ph.D. student at the University of Illinois Urbana-Champaign, where I am advised by Prof. Jian Huang. I build systems that make LLM agents more accessible by improving the efficiency of agent serving systems and driving down per-token inference cost.
I earned my B.S. in Computer Science at the University of Science and Technology of China (USTC).
π° Vegas: Self-Speculative Decoding with Verification-Guided Sparse Attention
Yikang Yue, Yuqi Xue, Jian Huang
ICML 2026
Preprint
Code
Poster
Vegas accelerates long-context LLM inference by reusing the full-attention
computation already available in the verification stage to guide sparse attention
in subsequent drafting steps.
π² CLAP: An FPGA CPU for the National Computer System Capability Challenge
With Zichen Zhang, Songxiao Guo, and Zirui Ma
Code
A LoongArch CPU implemented in Verilog with a dual-issue, eight-stage pipeline,
set-associative I/D caches, a 32-entry fully-associative TLB, and an efficient
branch predictor.
University of Illinois Urbana-Champaign, Champaign, IL
Ph.D. Student, advised by Prof. Jian Huang
Aug. 2024 – Present
University of Science and Technology of China, Hefei, China
B.S. in Computer Science
Sep. 2020 – Jun. 2024
Guo Moruo Scholarship, 2024
National Scholarship (China), 2021
2nd Place, National Student Computer System Capability Challenge – LoongArch Group, 2022
My Chinese name is δΉδΊ¦εΊ·, which means "Happy and Healthy".
When I'm not chasing TFLOPS, you can usually find me reading about
computer architecture history, exploring Urbana and Champaign on foot, or cooking
something from home.