GZGuangyi Zhang

Ph.D. Candidate · Zhejiang University

Guangyi Zhang 张光义

I am a researcher working at the intersection of semantic communication, efficient multimodal AI, and edge–cloud intelligence.

Explore my research → Download CV ↓

01 · About

Research with a systems perspective.

My work asks a simple question: how can intelligent models communicate, adapt, and reason under real-world resource constraints?

I am pursuing my Ph.D. at the College of Information Science and Electronic Engineering, Zhejiang University, advised by Prof. Yunlong Cai and Prof. Guanding Yu. From 2024 to 2025, I was a visiting Ph.D. researcher at King’s College London, advised by Prof. Osvaldo Simeone.

My research spans learned source–channel coding, multimodal semantic systems, adaptive edge–cloud inference, and prediction-powered monitoring—from statistical foundations to efficient implementation.

NowExploring efficient VLM/LLM inference, tool-using visual agents, and trustworthy monitoring for deployed AI.

02 · Research

Three connected directions

Communication-efficient learning—from the physical channel to multimodal models and distributed AI systems.

01
≋

Semantic Communication

Learned image transmission, multimodal task-oriented communication, secure semantic coding, and channel-adaptive systems.

  • JSCC
  • Multimodal coding
  • Adaptive rate
02
◌

Efficient Multimodal AI

Long-context VLM/LLM inference, hybrid linear attention, visual spatial reasoning, and tool-using agents.

  • VLM / LLM
  • Inference systems
  • Agents
03
◎

Edge–Cloud Intelligence

Adaptive speculative decoding, communication-aware acceleration, and statistically reliable model monitoring.

  • Speculative decoding
  • Risk monitoring
  • Edge AI

03 · Publications

Selected work

Prediction-powered risk monitoring framework

ICML 2026 · CCF-A

Prediction-Powered Risk Monitoring of Deployed Models for Detecting Harmful Distribution Shifts

Guangyi Zhang, Yunlong Cai, Guanding Yu, Osvaldo Simeone

Risk monitoringPrediction-powered inferenceDistribution shift

Semi-supervised monitoring that detects harmful shifts with finite-sample false-alarm guarantees.

Paper ↗

Unified multimodal semantic communication architecture2026 Stephen O. Rice Prize

IEEE TCOM · 2024

A Unified Multi-Task Semantic Communication System for Multimodal Data

Guangyi Zhang, Qiyu Hu, Zhijin Qin, Yunlong Cai, Guanding Yu, Xiaoming Tao

MultimodalTransformerSemantic communication

A shared Transformer system for transmitting text, image, and speech across multiple downstream tasks.

Paper ↗Code ↗Award ↗

Hierarchical VAE image transmission architecture

AAAI 2025 · CCF-A

Learned Image Transmission with Hierarchical Variational Autoencoder

Guangyi Zhang, Hanlei Li, Yunlong Cai, Qiyu Hu, Guanding Yu, Runmin Zhang

Learned image transmissionHierarchical VAEJSCC

Hierarchical latent representations and adaptive symbol allocation for robust wireless image transmission.

Paper ↗Code ↗

Adaptive edge-cloud speculative decoding

IEEE COMML · 2026

Quantize-Sample-and-Verify: LLM Acceleration via Adaptive Edge-Cloud Speculative Decoding

Guangyi Zhang, Yunlong Cai, Guanding Yu, Petar Popovski, Osvaldo Simeone

Speculative decodingEdge–cloud AIQuantization

Distribution-preserving quantization and adaptive draft scheduling for bandwidth-constrained edge–cloud inference.

Paper ↗Code ↗

Digital joint coding modulation architecture

IEEE TCOM · 2024

From Analog to Digital: Multi-Order Digital Joint Coding-Modulation for Semantic Communication

Guangyi Zhang, Pujing Yang, Yunlong Cai, Qiyu Hu, Guanding Yu

Digital codingLearned modulationSemantic communication

A digital semantic transmission framework learning coding and modulation across constellation orders.

Paper ↗Code ↗

Progressive learned image transmission

IEEE TCCN · 2025

Progressive Learned Image Transmission for Semantic Communication Using Hierarchical VAE

Guangyi Zhang, Hanlei Li, Yunlong Cai, Qiyu Hu, Guanding Yu, Zhijin Qin

Progressive codingHierarchical VAEAdaptive bandwidth

Progressive semantic features support graceful reconstruction across changing bandwidth budgets.

Paper ↗Code ↗

04 · Experience

Education & research experience

My academic training and visiting research experience across Zhejiang University and King’s College London.

2022 — Present

Doctoral study

Ph.D. in Information & Communication Engineering

Zhejiang University · Hangzhou, China

Research on semantic communication, learned image transmission, multimodal coding, and edge–cloud intelligence.

Semantic communicationMultimodal AIWireless learning
Dec 2024 — Nov 2025

Visiting Ph.D. researcher

Signal Processing & Networked AI

King’s College London · London, UK

Worked on communication-aware speculative decoding and prediction-powered monitoring for efficient and reliable deployed AI, in collaboration with Prof. Osvaldo Simeone.

LLM systemsSpeculative decodingRisk monitoring
2018 — 2022

Undergraduate study

B.Eng. in Information Engineering

Zhejiang University · Hangzhou, China

Outstanding Graduate

05 · Honors

Awards & recognition

Recognition for research impact, academic excellence, and engineering practice.

2026

IEEE Communications Society Stephen O. Rice Prize

Best paper published in IEEE Transactions on Communications · Top 0.1%

↗
2025

National Scholarship for Ph.D. Students

National Scholarship · Top 2%

2025

CAST Young Elite Scientists Sponsorship — Doctoral Program

China Association for Science and Technology · Top 1%

2024

Huawei Elite Cup Algorithm Competition

East China Regional Third Prize

2023 · 2024

Outstanding Graduate Student

Zhejiang University · Two consecutive years

2022

Outstanding Graduate

Zhejiang University

06 · Service

Academic community

I review for leading journals and conferences in communications and signal processing.

01

Journal reviewer

IEEE JSAC · TWC · TCOM · TCCN · WCL · SPL · COMML · TVT

02

Conference reviewer

IEEE GLOBECOM · ICC · WCNC · VTC · PIMRC · AAAI · ACM MM

Let’s connect

zhangguangyi@zju.edu.cn →