About Me

Hi, I am a Ph.D. candidate at the Australian Institute for Machine Learning (AIML), the University of Adelaide, supervised by A/Prof. Qi Wu and Dr. Yicong Hong. I am a member of the V3A Lab. Prior to my Ph.D. studies, I received my Master’s degree from the Australian National University supervised by Prof. Stephen Gould.

In 2025, I interned at Adobe Research, where I worked with Dr. Yicong Hong, Dr. Chongjian Ge, Dr. Hao Tan, and Dr. Tianyu Wang. I am currently a Research Intern on the Qwen Team. I co-lead Qwen-RobotNav with Dr. Jiazhao Zhang and have also contributed to Qwen-Image-3.0 and Qwen-RobotWorld.

I build explainable and embodied AI systems for autonomous agents that can perceive, reason, and navigate the physical world. My work focuses on integrating perception, reasoning, and long-horizon memory into a unified embodied intelligence framework that enables continuous learning and interpretable decision-making. I develop such systems by leveraging large multimodal reasoning models as decision policies, together with generative world models that support interaction, learning, and long-term evolution.

Some topics that I currently focus on include:

World Modeling and Simulation with Scalable Image/Video Generation: Qwen-Image-3.0, Qwen-RobotWorld, LightMover
Large Embodied Reasoning Models with Context Management and Agentic Learning: NavGPT, NavGPT-2
Embodied Navigation Foundation Models with Sim2Real Transferability: Qwen-RobotNav, NaVid, NavFoM

News

2026.08.05 We’re excited to celebrate the release of Qwen-Image-3.0! Congratulations to the whole team, and I’m delighted to work with so many talented people. Many thanks to Arena.ai for featuring Qwen-Image-3.0-Pro at #5 with 1,263 points on the Text-to-Image Arena, up from #15 and 1,191 points for Qwen-Image-2.0-Pro. [Leaderboard]
2026.06.16 We released Qwen-RobotNav, an agentic embodied navigation system we build at Qwen. I also had fun with the demo shoot, where we deployed the model on a robot dog and had it follow us around. You can see it in the blog. [Project]
2026.04.06 VLN-MME is accepted to ACL 2026. Congratulations to Xunyi for publishing his first paper!
2026.02.20 LightMover is accepted to CVPR 2026 and SAR to CVPR 2026 Findings. Thanks to all my mentors and collaborators at Adobe!
2026.01.05 I joined Qwen as a Research Intern working on VLA and VL post-train. Super excited to learn and work with the team!
2025.06.25 SAME is accepted to ICCV 2025. Thanks to all collaborators.
2025.04.14 I joined Adobe Research as a Research Intern working on text to video generation. Excited to work with the team at San Jose!
2025.01.27 One paper is accepted to ICRA 2025. Congratulations to Zerui!
2024.07.11 We are thrilled to see that @GoogleDeepMind shares the same perspective as our previous work NavGPT on instruction-following navigation agents and build fascinating robots based on Gemini 1.5 Pro! [Details]
2024.07.01 NavGPT-2 is accepted to ECCV 2024! Thanks to all collaborators.
2024.05.14 NaVid is accepted to RSS 2024! Congratulations to Jiazhao, Kunyu and Rongtao!
2023.12.09 Two papers are accepted to AAAI 2024. Congratulations and thanks to all collaborators.

Research

Document

Qwen-RobotNav: A Scalable Navigation Model Designed for an Agentic Navigation System

Jiazhao Zhang^*†, Gengze Zhou^*†, Hale Yin^*, et al.

Qwen Technical Report, 2026

One model for instruction following, PointNav, ObjectNav, target tracking, autonomous driving, and embodied question answering, with a configurable observation protocol and zero-shot real-world deployment.

LightMover: Generative Light Movement with Color and Intensity Controls

Gengze Zhou, Tianyu Wang, Soo Ye Kim, Zhixin Shu, Xin Yu, Yannick Hold-Geoffroy, Sumit Chaturvedi, Qi Wu, Zhe Lin, Scott Cohen

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026

Rethinking Training Dynamics in Scale-wise Autoregressive Generation

Gengze Zhou, Chongjian Ge, Hao Tan, Feng Liu, Yicong Hong

Findings of IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026

SAME: Learning Generic Language-Guided Visual Navigation with State-Adaptive Mixture of Experts

Gengze Zhou, Yicong Hong, Zun Wang, Chongyang Zhao, Mohit Bansal, Qi Wu

International Conference on Computer Vision (ICCV), 2025

NavGPT-2: Unleashing Navigational Reasoning Capability for Large Vision-Language Models

Gengze Zhou, Yicong Hong, Zun Wang, Xin Eric Wang, Qi Wu

European Conference on Computer Vision (ECCV), 2024

NavGPT: Explicit Reasoning in Vision-and-Language Navigation with Large Language Models

Gengze Zhou, Yicong Hong, Qi Wu

AAAI Conference on Artificial Intelligence (AAAI), 2024

Qwen-Image-3.0: Rich Content, Authentic Details, Deep Knowledge

Qwen Team

Built for useful visual work with prompts up to 4.5K tokens, legible text down to 10 px, native rendering in 12 languages, knowledge-rich generation, and realistic interfaces.