About
I'm a first-year PhD student at the School of Computer Science, University of Sydney, supervised by Prof. Chang Xu. Before starting my PhD in October 2025, I spent over three years as an Algorithm Researcher at Huawei Noah's Ark Lab (Feb 2022 – Sep 2025), where I worked on efficient vision models, diffusion-based image generation, and large vision-language models. I received my B.Eng. from Taiyuan University of Technology in 2021.
My past work has covered efficient object detection, structure-aware diffusion models for low-level vision, and positional encoding for large vision-language models. Currently, my research focuses on World Models, Vision-Language Models, and Agents — I'm hoping to explore, through the two modalities of vision and language, how models can learn to think and understand the world.
Research Interests
I'm mainly interested in World Models, Vision-Language Models, and Agents — hoping to explore, through the two directions of vision and language, how models can learn to think and understand the world.
News
- 05/2026 Circle-RoPE accepted by ICML 2026.
- 03/2026 Released Sibyl System, a fully autonomous AI research system with 20+ agents.
- 03/2026 Reached 1,000+ citations on Google Scholar.
- 12/2025 PocketLLM: Ultimate Compression of Large Language Models via Meta Networks accepted by AAAI 2026.
- 10/2025 Started my PhD at the University of Sydney, supervised by Prof. Chang Xu.
- 05/2025 Released Circle-RoPE, a cone-like decoupled rotary positional embedding for large vision-language models.
- 02/2025 DiC: Rethinking Conv3×3 Designs in Diffusion Models accepted by CVPR 2025.
- 02/2024 Released SAM-DiffSR and DenseMamba on arXiv.
- 09/2023 Gold-YOLO and Species196 accepted by NeurIPS 2023.
- 02/2022 Joined Huawei Noah's Ark Lab as an Algorithm Researcher.
Education
- 2025.10 – now Ph.D. in Computer Science University of Sydney · Advisor: Prof. Chang Xu
- 2017 – 2021 B.Eng. in Software Engineering Taiyuan University of Technology
Experience
- 2022.02 – 2025.09 Algorithm Researcher Huawei Noah's Ark Lab
Projects
A fully autonomous AI research system with 20+ specialized agents that orchestrate end-to-end ML research — from literature survey and hypothesis generation, through GPU experiment execution, to conference-ready paper writing — with zero human intervention. Built on a dual-loop architecture: inner loops refine individual projects, while outer loops let the system learn and evolve from past research cycles.
Publications
1,073 citations · h-index 8 · i10-index 8 — full list on Google Scholar. * denotes equal contribution.
Services
- Conference Reviewer: CVPR, ICML, NeurIPS, ICLR
- Journal Reviewer: IEEE T-PAMI