Jianqiao Zheng

Jianqiao Zheng

Lecturer

I am a Lecturer in the School of Artificial Intelligence at China University of Geosciences (Beijing). I received my PhD from the Australian Institute for Machine Learning (AIML), The University of Adelaide. My research interests include 3D and geometric vision, point clouds and neural fields, Vision Transformers, attention mechanisms, and positional encoding.

Professional Experience

  • 2026 - Present, Lecturer, School of Artificial Intelligence, China University of Geosciences (Beijing), China.
  • 2025.07 - 2025.12, PhD Researcher, The University of Adelaide, Australia.
  • 2019.08 - 2020.05, Research Assistant Intern, CI2CV Lab, Carnegie Mellon University, U.S.

Education

  • 2021.03 - 2025.06, The University of Adelaide The University of Adelaide, Australian Institute for Machine Learning (AIML), PhD. Supervisor: Prof. Simon Lucey.
  • 2017.09 - 2019.05, University of Michigan University of Michigan, Ann Arbor, M.S. in Computer Vision.
  • 2013.09 - 2017.05, Harbin Institute of Technology Harbin Institute of Technology, Honors School, B.S. in Automation.

Publications

Published Papers

[1] Jianqiao Zheng, Xueqian Li, Hemanth Saratchandran, Simon Lucey.
Structured Initialization for Attention in Vision Transformers.
NeurIPS 2025. [Paper]

[2] Jianqiao Zheng, Xueqian Li, Sameera Ramasinghe, Simon Lucey.
Robust Point Cloud Processing through Positional Embedding.
3DV 2024. [Paper]

[3] Xueqian Li, Jianqiao Zheng, Francesco Ferroni, Jhony Kaesemodel Pontes, Simon Lucey.
Fast Neural Scene Flow.
ICCV 2023. [Paper]

[4] Jianqiao Zheng, Sameera Ramasinghe, Xueqian Li, Simon Lucey.
Trading Positional Complexity vs. Deepness in Coordinate Networks.
ECCV 2022. [Paper]

[5] Jianqiao Zheng, Sameera Ramasinghe, Simon Lucey.
Rethinking Positional Encoding.
arXiv 2021. [Paper]

[6] Yiping Ji, James Martens, Jianqiao Zheng, Ziqin Zhou, Peyman Moghadam, Xinyu Zhang, Hemanth Saratchandran, Simon Lucey.
Cutting the Skip: Training Residual-Free Transformers.
ICLR 2026. [Paper]

Preprints

[7] Hemanth Saratchandran, Jianqiao Zheng, Yiping Ji, Wenbo Zhang, Simon Lucey.
Rethinking Attention: Polynomial Alternatives to Softmax in Transformers.
arXiv / under submission. [Paper]

[8] Jianqiao Zheng, Cameron Gordon, Yiping Ji, Hemanth Saratchandran, Simon Lucey.
From Tables to Signals: Revealing Spectral Adaptivity in TabPFN.
arXiv / under submission. [Paper]

Projects

ViT initialization
Structured ViT initialization

Structured ViT Initialization
NeurIPS 2025 work on injecting useful visual inductive bias into Vision Transformers through structured initialization while preserving scalable learning behavior.
[Paper]

3D vision
Robust point cloud processing

Robust Point Cloud Processing
3DV 2024 project on improving point cloud robustness through positional embeddings, with theoretical analysis of bandwidth control and empirical evaluation under noise and distribution shifts.
[Project] [Code] [Paper]

coordinate networks
Coordinate networks

Trading Positional Complexity vs. Deepness in Coordinate Networks
An ECCV 2022 project studying the trade-off between positional encoding complexity and network depth for coordinate-based neural representations.
[Project] [Code] [Paper]

Teaching

  • 2023 - 2025, The University of Adelaide, Teaching Assistant, COMP SCI 3315 Computer Vision.
  • 2025, The University of Adelaide, Teaching Assistant, MATHS 1004 Mathematics for Data Science I.
  • 2018, University of Michigan, Teaching Assistant, SI 670 Applied Machine Learning.

Honors

  • Dean’s Commendation for Doctoral Thesis Excellence.
  • Gold Prize, China International College Students’ Innovation Competition 2024, International Track Grand Final.