Wenzhang Sun 孙文章

I am a postdoctoral researcher at Li Auto through a joint program with Tsinghua University. My research focuses on video generation, controllable generative models, and compact spatiotemporal representations.

I am particularly interested in advancing these capabilities toward interactive world models and Physical AI.

I received my Ph.D. from Beijing Institute of Technology, where I worked on 3D human modeling and neural reconstruction.

news

Recent updates

  • MUSE accepted to the ACM Multimedia Workshop on Agentic Multimodal Intelligence 2026.
  • SeMo accepted to ACM MM 2026 (Oral).
  • Holo-World released on arXiv with a project page.
  • RiO-DETR accepted to ECCV 2026 (Oral).
  • PREX released on arXiv with a project page.
  • RiO-DETR released as an arXiv preprint.
  • MUSE released as an arXiv preprint.
  • DrivingScene and PAGS accepted to ICASSP 2026.
  • FaceVid-1K accepted to ICCV 2025.
  • MoEE accepted to CVPR 2025.

selected work

Current focus

Controllable video generation and representation

My recent work explores geometry-guided video editing, joint camera, object and weather control, compact video representations, and multimodal generation and evaluation.

research

Publications

Publications and preprints, grouped by year. Download BibTeX.

2026

MUSE: A Multi-agent Framework for Unconstrained Story Envisioning via Closed-Loop Cognitive Orchestration

W. Sun, Z. Wang, Z. Hu, C. Wang, H. Li, W. Chen.

ACM Multimedia Workshop on Agentic Multimodal Intelligence 2026
paperproject
MUSE long-form audio-visual storytelling teaser

SeMo: A Self-supervised Motion Latent for Portrait Video Generation

Q. Zhang, C. Wu, W. Sun, H. Liu, et al.

ACM MM 2026 Oral
paper
SeMo motion representation and generation overview

Holo-World: Unified Camera, Object and Weather Control for Video World Model

Xiangchen Yin, Wenzhang Sun, Jiahui Yuan, Zijie Liu, Yinda Chen, Wei Li, Dachun Kai, Chunfeng Wang, Xiaoyan Sun.

arXiv 2026
paperproject
Holo-World unified camera, object, and weather control teaser

RiO-DETR: DETR for Real-time Oriented Object Detection

Z. Hu, Y. Zhao, Y. Peng, W. Sun, et al.

ECCV 2026 Oral
paper
RiO-DETR speed and accuracy comparison

Beyond Logits: Coherent Hallucination Mitigation via Attention Contrastive Decoding

Yujia Chen, Rui Sun, Wangkai Li, Huayu Mai, Bingzhou Wang, Zhangyu He, Aibing Li, Wenzhang Sun, Tianzhu Zhang.

ICML 2026
paper
Attention Contrastive Decoding overview

Beyond Blind Noising: Disentangled Visual Rectification for Hallucination Mitigation in MLLMs

Yujia Chen, Rui Sun, Zhaoyang Li, Wangkai Li, Huayu Mai, Bingzhou Wang, Aibing Li, Wenzhang Sun.

ICML 2026
paper
Disentangled Visual Rectification overview

Preserve, Reveal, Expand: Faithful 4D Video Editing with Region-Aware Conditioning

Zhangchi Hu, Wenzhang Sun, Xiangchen Yin, Jiahui Yuan, Chunfeng Wang, Hao Li, Kun Zhan, Xiaoyan Sun. Project Leader.

arXiv 2026
paperproject
PREX and PREBench overview

Stable Curves, Unstable Items: Item-Level Scaling Heterogeneity in Video LLMs

Wenzhang Sun, Chunfeng Wang, Xiangchen Yin, Yujia Chen, Hao Li, Kun Zhan.

arXiv 2026
code
Video LLM item-level scaling overview

DrivingScene: A Multi-Task Online Feed-Forward 3D Gaussian Splatting Method for Dynamic Driving Scenes

Q. Hou, W. Sun, C. Zeng, C. Wang, H. Li, J. Cui.

ICASSP 2026
paper
DrivingScene depth and optical-flow predictions

PAGS: Priority-Adaptive Gaussian Splatting for Dynamic Driving Scenes

A. Ying, W. Sun, C. Zeng, C. Wang, H. Li, J. Cui.

ICASSP 2026
paper
PAGS dynamic driving-scene reconstruction comparison

2025

UniCP: A Unified Caching and Pruning Framework for Efficient Video Generation

W. Sun, Q. Hou, D. Di, J. Yang, Y. Ma, J. Cui.

MMAsia 2025
paper
UniCP video generation comparison

DeCo-VAE: Learning Compact Latents for Video Reconstruction via Decoupled Representation

X. Yin, J. Yuan, Z. Hu, W. Sun, et al.

arXiv 2025
paper
DeCo-VAE decoupled video representation overview

TV-3DG: Mastering Text-to-3D Customized Generation with Visual Prompt

Jiahui Yang, Donglin Di, Baorui Ma, Jianxun Cui, Xun Yang, Yongjia Ma, Wenzhang Sun, Wei Chen, Zhou Xue, Meng Wang, Yebin Liu.

TPAMI 2025
paper
TV-3DG customized 3D generation examples

FaceVid-1K: A Large-Scale High-Quality Multiracial Human Face Video Dataset

D. Di, H. Feng, W. Sun, Y. Ma, et al.

ICCV 2025
project
FaceVid-1K dataset overview

Hi-VAE: Efficient Video Autoencoding with Global and Detailed Motion

H. Liu*, W. Sun*, Q. Zhang, D. Di, et al.

arXiv 2025
paper
Hi-VAE compact video latent comparison

ChronoTailor: Harnessing Attention Guidance for Fine-Grained Video Virtual Try-On

J. Wang*, W. Sun*, M. Li, Y. Zheng, et al.

arXiv 2025
paper
ChronoTailor video virtual try-on results

MoEE: Mixture of Emotion Experts for Audio-Driven Portrait Animation

H. Liu, W. Sun, D. Di, S. Sun, J. Yang, C. Zou, H. Bao.

CVPR 2025
paper
MoEE basic and compound emotion control examples

2024 and earlier

UniAvatar: Taming Lifelike Audio-Driven Talking Head Generation with Comprehensive Motion and Lighting Control

W. Sun, X. Li, D. Di, Z. Liang, Q. Zhang, H. Li, W. Chen, J. Cui.

arXiv 2024
paper
UniAvatar motion and lighting control examples

Neural Reconstruction of Relightable Human Model from Monocular Video

W. Sun, Y. Che, H. Huang, Y. Guo.

ICCV 2023
paper
Relightable human reconstruction teaser

Estimating 3D Body Mesh without SMPL Annotations via Alternating Successive Convex Approximation

W. Sun, L. Wang, S. Ma, Q. Ma.

CVIU 2022
Alternating shape and pose estimation architecture

professional service

Peer Review

Conference Reviewer

  • NeurIPS
  • ICLR
  • CVPR
  • ECCV
  • AAAI
  • ICASSP

Journal Reviewer

IEEE Transactions on Multimedia (TMM)