Guangxuan Xiao   肖光烜

I am a first-year Ph.D. student at MIT EECS, advised by Prof. Song Han. I graduated from the Department of Computer Science and Technology at Tsinghua University in 2022, advised by Prof. Zhiyuan Liu and Prof. Jidong Zhai. I have been very fortunate to work with Prof. Jure Leskovec as a summer intern in 2020, with Prof. Jiajun Wu and Prof. Leslie Pack Kaelbling through the Stanford UGVR Program.

Email  /  Google Scholar  /  Github

profile photo
Research

My research interests are efficient algorithm and systems for real-world machine learning applications, especially foundation models.

Offsite-Tuning: Transfer Learning without Full Model
Guangxuan Xiao, Ji Lin, Song Han
arXiv
[paper] [code]

ReFresh: Reducing Memory Access from Exploiting Stable Historical Embeddings for Graph Neural Network Training
Kezhao Huang, Haitian Jiang, Minjie Wang, Guangxuan Xiao, David Wipf, Xiang Song, Quan Gan, Zengfeng Huang, Jidong Zhai, Zheng Zhang
arXiv
[paper]

SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models
Guangxuan Xiao*, Ji Lin*, Mickael Seznec, Julien Demouth, Song Han
arXiv
[paper] [code]

Sparse and Local Hypergraph Reasoning Networks
Guangxuan Xiao, Leslie Pack Kaelbling, Jiajun Wu, Jiayuan Mao
LOG 2022
[paper]

Red Alarm for Pre-trained Models: Universal Vulnerability to Neuron-Level Backdoor Attacks
Zhengyan Zhang*, Guangxuan Xiao*, Yongwei Li, Tian Lv, Fanchao Qi, Zhiyuan Liu, Yasheng Wang, Xin Jiang, Maosong Sun
ICML 2021 AdvML Workshop
[paper]

Projects
Hardware Router on RISC-V CPU
Guangxuan Xiao, Tian Lv, Zizhao Zhang

We implemented a 5-stage pipeline RV32I CPU on FPGA from scratch. Then we implemented a tiny OS and the RIPv2 routing algorithm on the CPU, getting a high-performance IPv4 router.
[code][doc]

Realistic Rendering based on Path Tracing and Photon Mapping

I implemented a realistic rendering engine using path tracing and stochastic progressive photon mapping algorithms, boosted with bounding boxes, hierarchical KD-Tree, CUDA, and OpenMP.
[code]

Recognizing, Detecting and Generating Oracle Characters with Deep Learning
Guangxuan Xiao, Yongwei Li, Tian Lv

We proposed a well-organized oracle characters dataset and a novel data augmentation method to detect oracle characters from raw oracle bones. We implemented generative adversarial networks (GANs) to generate and complete oracle characters.

Experience
Massachusetts Institute of Technology
2022.08 - Present

Ph.D student in Computer Science
Advisor: Prof. Song Han
Tsinghua University
2018.08 - 2022.07

B.Eng. in Computer Science
B.Econ. in Economics (Second Major)
Advisor: Prof. Zhiyuan Liu, Prof. Jidong Zhai
Stanford University
2020.07 - 2021.06

Visiting Research Student
Advisor: Prof. Jure Leskovec
Mentor: Jiaxuan You
Stanford University
2021.06 - 2021.11

Visiting Research Student through the UGVR program
Advisor: Prof. Jiajun Wu, Prof. Leslie Pack Kaelbling
Mentor: Jiayuan Mao
Honors & Awards
  • Tsinghua Comprehensive Excellence Scholarship, 2019
  • National Scholarship, 2020
  • Boeing Scholarship, 2021
  • SenseTime Scholarship, 2021
  • Tsinghua "Future Scholar" Scientific Research Grant (30000$), 2021
  • Contemporary Undergraduate Mathematical Contest in Modeling, 1st Prize, 2020
  • Beijing "Challenge Cup" Academic Science and Technology Competition, 1st Prize, 2020
Miscellaneous

I love to play soccer. I was the captain and striker of my department soccer team.
I also love to play table tennis, Go (Weiqi), and piano. Beethoven's works are my favorite.


Template adapted from Jon Barron.