Genghan Zhang

PhD Student Department of Computer Science Stanford University Office: 478 Gates zgh23 [at] stanford [dot] edu Curriculum Vitae

I am a computer science PhD student at Stanford University advised by Professor Kunle Olukotun. I also worked with Professor Fredrik Kjolstad on sparse tensor algebra compilation and Professor Azalia Mirhoseini on efficient sparse large language models.

Research: My research interests lie in better programming models and systems for domain-specific architectures. I am also interested in optimizing GPU kernels for emerging applications, including sparse and recurrent neural networks.

I graduated from Tsinghua University in 2023 with a bachelor degree in Eletronic Engineering. At Tsinghua, I did research at NICS-EFC Lab on effcient sparse tensor algebra for GPU and IDEAL Lab on kernel architecture search.

Selected Publications

AccelOpt: A Self-Improving LLM Agentic System for AI Accelerator Kernel Optimization Genghan Zhang, Shaowei Zhu, Anjiang Wei, Zhenyu Song, Allen Nie, Zhen Jia, Nandita Vijaykumar, Yida Wang, and Kunle Olukotun Preprint, November 2025
Adaptive Self-improvement LLM Agentic System for ML Library Development Genghan Zhang, Weixin Liang, Olivia Hsu, and Kunle Olukotun ICML 2025 Poster (acceptance rate: 26.9%); ICML 2025 ES-FoMo Workshop, July 2025 ICLR 2025 DL4C Workshop Best Paper Award (2/63)
Learning to (Learn at Test Time): RNNs with Expressive Hidden States Yu Sun†, Xinhao Li†, Karan Dalal†, Jiarui Xu, Arjun Vikram, Genghan Zhang, Yann Dubois, Xinlei Chen, Xiaolong Wang, Sanmi Koyejo, Tatsunori Hashimoto, and Carlos Guestrin † co-first author ICML 2025 Spotlight (acceptance rate 2.59%), July 2025
Compilation of Modular and General Sparse Workspaces Genghan Zhang, Olivia Hsu, and Fredrik Kjolstad PLDI 2024 (acceptance rate 27.64%), June 2024
CATS: Context-Aware Thresholding for Sparsity in Large Language Models Donghyun Lee†, Jeyong Lee†, Genghan Zhang, Mo Tiwari, and Azalia Mirhoseini † co-first author COLM 2024 (acceptance rate 28.86%), October 2024
Sgap: Towards Efficient Sparse Tensor Algebra Compilation for GPU Genghan Zhang, Yuetong Zhao, Yanting Tao, Zhongming Yu, Guohao Dai, Sitao Huang, Yuan Wen, Pavlos Petoumenos, and Yu Wang CCF Transactions on High Performance Computing, 2023

Blogs

One Agent, Any Chip: Expert-free, Self-Improving Kernel Optimization Genghan Zhang December 22, 2025

Why "LLM for Compiler" Is a Reasonable Research Question Genghan Zhang May 27, 2025

Adaptive Self-improvement LLM Agentic System for ML Library Development Genghan Zhang , Weixin Liang , Olivia Hsu , and Kunle Olukotun March 01, 2025

Talks

Self-improving LLM Agentic Systems for Programming AI Accelerators November 2025 UCSD MLSys Seminar
Adaptive Self-improvement LLM Agentic System for ML Library Development June 2025 Stanford SEAMS Retreat
Compilation of Modular and General Sparse Workspaces April 2025 Stanford Software Research Lunch
Adaptive Self-improvement LLM Agentic System for ML Library Development April 2025 Uber Programming Systems Group (PSG) Invited Talk Series