Yuanzhi Liang

Mail: liangyzh18 [at] outlook [dot] com

Bilibili zhihu Xiaohongshu

About Me

I am a research scientist specializing in generative AI at the Institute of Artificial Intelligence (TeleAI), China Telecom. I received my Ph.D. from the University of Technology Sydney in 2024, advised by Dr. Linchao Zhu and Prof. Yi Yang.

I received a Master's degree from Xi'an Jiaotong University in 2020 and was a member of the SMILES LAB, advised by Prof. Xueming Qian and Prof. Li Zhu.

My research focuses on improving machine intelligence that learns and generalizes from visual content, with particular interests in post-training of generative visual systems, video generation, and world models. I am also interested in theoretical innovations in generative models.

I am always looking for highly motivated research interns and long-term collaborators. We currently have multiple positions available, focusing on, but not limited to, multimodal large models, video generation/editing, and 3D generation. If you are interested in exploring these areas or discussing potential research collaborations, please feel free to contact me via email. (Applicants for internships are encouraged to include your CV.)

Work Experience

  • Jul 2021 - Dec 2021, Alibaba DAMO Academy
    • Research intern working on virtual human synthesis.

  • Jul 2020 - Jul 2021, Baidu Research
    • Research intern working on visual knowledge embedding, object recognition, and multi-modal representation.

  • Mar 2020 - Jun 2020, JD AI Research
    • Research intern working on product recognition.

  • Aug 2018 - Jun 2019, JD AI Research
    • Research intern working on visual-language representation learning.

Selected Honors

  • First place in AliProducts Challenge @ CVPR 2020 the RetailVision workshop.
  • First place in iMat Product Competition @ CVPR 2019 FGVC6 workshop.
  • First place in in Fieldguide Challenge: Moths & Butterflies @ CVPR 2019 FGVC6 workshop.
  • Second place in iFood Competition @ CVPR 2019 FGVC6 workshop.
  • Second place in iMet2020 Fine-grained Attributes Classification Competition @ CVPR 2020 FGVC7 workshop.
  • Kaggle Silver Medal in Deepfake Detection Challenge 2020.

Important Preprints

  • Seeing What Matters: Visual Preference Policy Optimization for Visual Generation
    Ziqi Ni*, Yuanzhi Liang, Rui Li, Yi Zhou, Haibing Huang, Chi Zhang, Xuelong Li.
    Arxiv, 2025

  • Growing with the Generator: Self-paced GRPO for Video Generation
    Rui Li*, Yuanzhi Liang, Ziqi Ni, Haibing Huang, Chi Zhang, Xuelong Li.
    Arxiv, 2025

  • Learning What to Trust: Bayesian Prior-Guided Optimization for Visual Generation
    Ruiying Liu*, Yuanzhi Liang, Haibin Huang, Tianshu Yu, Chi Zhang.
    Arxiv, 2025

  • Free3D: 3D Human Motion Emerges from Single-View 2D Supervision
    Sheng Liu*, Yuanzhi Liang, Sidan Du.
    Arxiv, 2025

  • VAST 1.0: A Unified Framework for Controllable and Consistent Video Generation.
    Chi Zhang, Yuanzhi Liang, Xi Qiu, Fangqiu Yi, Xuelong Li.
    Arxiv, 2024

    • Selected Publications

  • Uni-Inter: Unifying 3D Human Motion Synthesis Across Diverse Interaction Contexts
    Sheng Liu*, Yuanzhi Liang, Jiepeng Wang, Sidan Du, Chi Zhang, Xuelong Li.
    Accepted by SIGGRAPH Asia 2025

  • Intersyn: Interleaved learning for dynamic motion synthesis in the wild
    Yiyi Ma*, Yuanzhi Liang, Xiu Li, Chi Zhang, Xuelong Li.
    Accepted by ICCV 2025

  • MAAL: Multimodality-Aware Autoencoder-based Affordance Learning for 3D Articulated Objects.
    Yuanzhi Liang, Xiaohan Wang, Linchao Zhu, Yi Yang.
    Accepted by ICCV 2023

  • A Simple Episodic Linear Probe Improves Visual Recognition in the Wild.
    Yuanzhi Liang, Linchao Zhu, Xiaohan Wang, Yi Yang.
    Accepted by CVPR 2022 (Score 1/2/2)

  • SEEG: Semantic Energized Co-speech Gesture Generation.
    Yuanzhi Liang, Qianyu Feng, Linchao Zhu, Li Hu, Pan Pan, Yi Yang.
    Accepted by CVPR 2022

  • VrR-VG: Refocusing Visually-Relevant Relationships.
    Yuanzhi Liang, Yalong Bai, Wei Zhang, Xueming Qian, Li Zhu, Tao Mei.
    Accepted by ICCV 2019

  • IcoCap: Improving Video Captioning by Compounding Images.
    Yuanzhi Liang, Linchao Zhu, Xiaohan Wang, Yi Yang.
    Accepted by TMM 2023

  • Penalizing the Hard Example But Not Too Much: A Strong Baseline for Fine-Grained Visual Classification.
    Yuanzhi Liang, Linchao Zhu, Xiaohan Wang, Yi Yang.
    Accepted by TNNLS 2022

  • Towards Better Railway Service: Passengers Counting in Railway Compartment.
    Yuanzhi Liang, Zhu Li, Xueming Qian.
    Accepted by TCSVT 2020

  • Freelong: Training-free long video generation with spectralblend temporal attention.
    Yu Lu, Yuanzhi Liang, Linchao Zhu, Yi Yang.
    Accepted by NeurIPS 2024

  • Removing Raindrops and Rain Streaks in One Go.
    Ruijie Quan, Xin Yu, Yuanzhi Liang, Yi Yang.
    Accepted by CVPR 2021

  • Food and Ingredient Joint Learning for Fine-Grained Recognition.
    Chengxu Liu, Yuanzhi Liang, Yao Xue, Xueming Qian, Jianlong Fu.
    Accepted by TCSVT 2020

    • Note *: interns that I mentored.

    Journal Reviewer

  • Reviewer for TPAMI and TIP.

    • Conference Reviewer / Program Committee Member

  • Reviewer for ICCV, CVPR, ICLR, NeurIPS, ECCV, MM, AAAI, IJCAI, ICME, and CAAI.

    • Others

  • Member of MNBVC (Massive Never-ending BT Vast Chinese corpus).