Lijun Zhang

I am an Assistant Professor in Computer Science at Zhejiang University of Technology. Before joining ZJUT, I was a Postdoctoral Scientist at Amazon Robotics, where I worked on perception and world modeling for robotic systems.

I received my Ph.D. from the College of Information and Computer Sciences at the University of Massachusetts Amherst, advised by Prof. Hui Guan. Before that, I received my Master’s and Bachelor’s degrees from the School of Software Engineering at Tongji University, advised by Prof. Lin Zhang.

My research lies at the intersection of automated machine learning and computer vision. My previous work has focused on resource-efficient multi-task learning, particularly task relationship modeling, parameter sharing, and automatic architecture design, as well as applying diffusion models beyond image generation to image restoration, watermarking, parameter-space exploration, and visual world modeling. I am currently extending this line of research toward adaptive and embodied agents, with an emphasis on how agents acquire, represent, and reuse knowledge through interaction with new tasks and environments.

News

• [Aug. 27, 2026]: I joined the faculty of Zhejiang University of Technology as an Assistant Professor in Computer Science.
• [May. 2026]: Our work “Communication-Efficient Multi-Device Inference Acceleration for Transformer Models” has been accepted to ICML’26.
• [Sept. 2025]: Our work “Attacking all tasks at once using adversarial examples in multi-task learning” has been accepted to Neurocomputing.
• [June. 2025]: Our work “Reimagining Parameter Space Exploration with Diffusion Models” has been accepted to ICML’25 EXAIT.
• [April. 2025]: I have successfully defended my PhD Thesis on “Advanced Resource-Efficient Multi-Task Learning”.
• [Sept. 2024]: Our work “Attack-Resilient Image Watermarking Using Stable Diffusion” has been accepted to NeurIPS’24.
• [Sept. 2024]: Our work “Thinking Forward: Memory-Efficient Federated Finetuning of Language Models” has been accepted to NeurIPS’24.
• [May. 2024]: Start an internship at Advanced Technology Group (ATG) in Dolby Laboratories.
• [Feb. 2024]: Our work “GMorph: Accelerating Multi-DNN Inference via Model Fusion” has been accepted to EuroSys’24.
• [Sept. 2023]: Our work “Flow: Per-instance Personalized Federated Learning” has been accepted to NeurIPS’23.
• [Sept. 2023]: Receive the 2023 IBM PhD Fellowship Award.
• [May. 2023]: Start an internship in Amazon Robotics.
• [Sept. 2022]: Our paper “AutoMTL: A Programming Framework for Automating Efficient Multi-Task Learning” has been accepted to NeurIPS’22.
• [May. 2022]: Our paper “A Tree-Structured Multi-Task Model Recommender” has been accepted to AutoML’22.
• [Mar. 2022]: Our paper “Rethinking Hard-Parameter Sharing in Multi-Domain Learning” has been accepted to ICME’22.
• [May 2021]: Our paper “Reuse-Centric Kmeans Configuration” has been accepted to Information Systems.

Awards

• [2026] Gold Reviewer, ICML
• [2024] UMass CICS PhD Dissertation Writing Fellowship
• [2023] IBM Ph.D. Fellowship
• [2022] Scholar Award & Top Reviewer, NeurIPS
• [2022] Travel Grant Award, Conf-AutoML
• [2020] The Lori a. Clarke Scholarship, UMass
• [2019 & 2016] Best Graduate & Undergraduate Thesis, Tongji University
• [2019 & 2016] Outstanding Graduate in Shanghai
• [2018] National Scholarship for Graduate Students

Services

• Area Chair: ICLR 2027
• Conference Reviewer:
  • NeurIPS 2022-2026, ICML 2023-2026, ICLR 2023-2026, AAAI 2027
  • CVPR 2024-2025, ICCV 2025, ECCV 2026
• Journal Reviewer: TMLR Yearly Reviewer