Jiacheng Zhao

Associate Professor

ICT, CAS

zhaojiacheng@ict.ac.cn

Dr. Jiacheng Zhao is an Associate Professor and Master Advisor at State Key Lab of Processors, Institute of Computing Technology, Chinese Academy of Sciences (SKLP, ICT-CAS). I obtained my Ph.D. degree in July 2017 in ICT, CAS under the supervision of Professor Feng Xiaobing. I received my bachelor’s degree in computer science from Tianjin University in 2012.

Recently, my research focuses on building next-gen compiler infrastructure for domain specific accelerators, e.g. GPUs, AI Chips and network processors.

I’m looking for self-motivated students (Ph.D., M.S. and/or undergraduate students) to collaborate with me in the general area of compilers ranging from heterogeneous compiler construction and machine learning systems, to compiler assisted runtime systems. Send me an email if you are interested.

Research Interests

Compiler Construction
Heterogeneous Programming
Machine Learning Systems
Compiler Infrastructure for Domain Specific Architectures

Education

Ph.D., Institute of Computing Technology, Chinese Academy of Sciences, Computer Architecture (2012-2017). Supervised by Prof. Xiaobing Feng. Awarded National Scholarship for Graduate Students of China (2013&2016), Special Prize of President Scholarship (Xia Peisu Scholarship), Pacemaker to Merit Student of Chinese Academy of Sciences
B.Sc., Tianjin University, Computer Science and Technology (2008-2012)

news

Jul 16, 2026	Ave: Guiding Agentic GPU Optimization using Data-Flow Invariants accepted to SOSP 2026. July 2026.
May 05, 2026	Two papers on re-architecting GPU memory management through driver-runtime co-design accepted to OSDI 2026 and ICML 2026, with the ICML paper selected as a Spotlight (top 2.2%). May 2026.
May 05, 2026	Symbiotic MLLM Serving accepted to ISCA 2026. May 2026.
May 05, 2026	Our work on structured output as a control-plane vulnerability in LLM jailbreaking accepted to CCS 2026. May 2026.
Feb 03, 2026	T2T accepted to CGO 2026 and won the Distinguished Paper Award (3 out of 56 papers). Feb 2026.

selected publications

Ave: Guiding Agentic GPU Optimization using Data-Flow Invariants

Haohui Mai, Xiaoyan Guo, Xiangyun Ding, and 7 more authors

The 32nd ACM Symposium on Operating Systems Principles (SOSP 2026), to appear, Jul 2026

Bib

@article{pub29,
  title = {Ave: Guiding Agentic GPU Optimization using Data-Flow Invariants},
  author = {Mai, Haohui and Guo, Xiaoyan and Ding, Xiangyun and Li, Daifeng and Yu, Qiuchu and Guo, Chenzhun and Wang, Cong and Zhao, Jiacheng and Kozyrakis, Christos and Yuan, Binhang},
  journal = {The 32nd ACM Symposium on Operating Systems Principles (SOSP 2026), to appear},
  year = {2026},
  month = jul,
}

When Grammar Guides the Attack: Uncovering Control-Plane Vulnerabilities in LLMs with Structured Output

Shuoming Zhang, Jiacheng Zhao, Hanyuan Dong, and 9 more authors

The ACM Conference on Computer and Communications Security (CCS 2026), to appear, May 2026

Bib

@article{pub27,
  title = {When Grammar Guides the Attack: Uncovering Control-Plane Vulnerabilities in LLMs with Structured Output},
  author = {Zhang, Shuoming and Zhao, Jiacheng and Dong, Hanyuan and Xu, Ruiyuan and Li, Zhicheng and Zhang, Yangyu and Li, Shuaijiang and Wen, Yuan and Xia, Chunwei and Wang, Zheng and Feng, Xiaobing and Cui, Huimin},
  journal = {The ACM Conference on Computer and Communications Security (CCS 2026), to appear},
  year = {2026},
  month = may,
}

Symbiotic MLLM Serving: Dynamically Balancing Parallelism Across GPUs and Resources Within GPUs

Zhicheng Li, Jiacheng Zhao, Yangyu Zhang, and 10 more authors

The 53rd Annual International Symposium on Computer Architecture (ISCA 2026), to appear, May 2026

Bib

@article{pub23,
  title = {Symbiotic MLLM Serving: Dynamically Balancing Parallelism Across GPUs and Resources Within GPUs},
  author = {Li, Zhicheng and Zhao, Jiacheng and Zhang, Yangyu and Duan, Zhaolin and Liu, Xinyu and Li, Siqi and Zhang, Shuoming and Li, Shuaijiang and Yu, Donglin and Wen, Yuan and Xia, Chunwei and Shi, Xiyu and Cui, Huimin},
  journal = {The 53rd Annual International Symposium on Computer Architecture (ISCA 2026), to appear},
  year = {2026},
  month = may,
}

LEGO: An LLM-Enabled Hierarchical Optimizer for Tensor Computation Graphs with Structure-Aware Search and Compositional Synthesis

Ruiyuan Xu, Shuoming Zhang, Guangli Li, and 10 more authors

International Conference on Machine Learning (ICML 2026), to appear, May 2026

Bib

@article{pub26,
  title = {LEGO: An LLM-Enabled Hierarchical Optimizer for Tensor Computation Graphs with Structure-Aware Search and Compositional Synthesis},
  author = {Xu, Ruiyuan and Zhang, Shuoming and Li, Guangli and Yu, Qiuchu and Zhang, Rui and Zhang, Yangyu and Qian, Hao and Xia, Chunwei and Zhao, Jiacheng and Wang, Chenxi and Feng, Xiaobing and Xue, Jingling and Cui, Huimin},
  journal = {International Conference on Machine Learning (ICML 2026), to appear},
  year = {2026},
  month = may,
}

CONTINUUM: Restoring the Contiguous Tensor Abstraction Efficiently for Dynamic AI Workloads via Hardware Virtualization

Yangyu Zhang, Shuoming Zhang, Chunwei Xia, and 10 more authors

International Conference on Machine Learning (ICML 2026), Spotlight (top 2.2%), to appear, May 2026

Bib

@article{pub25,
  title = {CONTINUUM: Restoring the Contiguous Tensor Abstraction Efficiently for Dynamic AI Workloads via Hardware Virtualization},
  author = {Zhang, Yangyu and Zhang, Shuoming and Xia, Chunwei and Li, Shuaijiang and Li, Zhicheng and Xu, Ruiyuan and Yang, Zheming and Chen, Lei and Wen, Yuan and Li, Guangli and Feng, Xiaobing and Cui, Huimin and Zhao, Jiacheng},
  journal = {International Conference on Machine Learning (ICML 2026), Spotlight (top 2.2%), to appear},
  year = {2026},
  month = may,
}

A GPU Memory Allocator with Device-Side Page Table Materialization and Deferred TLB Coherence

Yangyu Zhang, Lei Chen, Chunwei Xia, and 12 more authors

USENIX Symposium on Operating Systems Design and Implementation (OSDI 2026), to appear, May 2026

Bib

@article{pub24,
  title = {A GPU Memory Allocator with Device-Side Page Table Materialization and Deferred TLB Coherence},
  author = {Zhang, Yangyu and Chen, Lei and Xia, Chunwei and Li, Shuaijiang and Zhang, Shuoming and Li, Zhicheng and Sun, Qianqi and Xiao, Jiawei and Xu, Ruiyuan and Chen, Ao and Li, Guangli and Feng, Xiaobing and Cui, Huimin and Wang, Chenxi and Zhao, Jiacheng},
  journal = {USENIX Symposium on Operating Systems Design and Implementation (OSDI 2026), to appear},
  year = {2026},
  month = may,
}

From Threads to Tiles: T2T, a Compiler for CUDA-to-NPU Translation via 2D Vectorization

Shuaijiang Li, Jiacheng Zhao, Ying Liu, and 11 more authors

The 24th ACM/IEEE International Symposium on Code Generation and Optimization (CGO 2026), Distinguished Paper Award, Feb 2026

Bib

@article{pub16,
  title = {From Threads to Tiles: T2T, a Compiler for CUDA-to-NPU Translation via 2D Vectorization},
  author = {Li, Shuaijiang and Zhao, Jiacheng and Liu, Ying and Zhang, Shuoming and Chen, Lei and Li, Yijin and Zhang, Yangyu and Li, Zhicheng and Zhou, Runyu and Shi, Xiyu and Xia, Chunwei and Wen, Yuan and Feng, Xiaobing and Cui, Huimin},
  journal = {The 24th ACM/IEEE International Symposium on Code Generation and Optimization (CGO 2026), Distinguished Paper Award},
  year = {2026},
  month = feb,
}