Zihan Su (苏梓瀚)

Hi! I am Zihan Su (Chinese name: 苏梓瀚), a master's student in Computer Technology at Tsinghua University, advised by Prof. Chun Yuan. I am currently a Foundation Model Research Intern at Joy Future Academy, working on world models and real-time interactive video generation.

My research interests include:

World Models & Real-Time Interactive Video Generation
Unified Multimodal Generation & Understanding
Post-Training for Text-to-Image Models

Open to full-time and internship opportunities in world models and real-time interactive video generation. Get in touch ↗ 寻找世界模型与实时交互式视频生成方向的正式及实习岗位，欢迎联系。

News

08/2026 🎉 EchoWM was released as a Technical Report.
07/2026 🎉 Joined Joy Future Academy as a Foundation Model Research Intern.
04/2026 🎉 Started an AIGC Application Internship in ByteDance's Jindouyun Talent Program.
01/2026 🎉 UniMRG was accepted to ICML 2026 as a first-author paper.
09/2025 🎉 Joined Alibaba as an Academic Research Intern.
09/2025 🎉 Safe-Sora was accepted to NeurIPS 2025 as a first-author paper.
09/2025 🏆 Received the Tsinghua Friend - Ordos Talent Scholarship.
06/2025 🎉 Joined Tencent WXG as an Application Intern.
01/2025 🎉 Joined Guangming Laboratory as an Academic Research Intern.
01/2025 🎉 TextureDiffusion was accepted as an ICASSP 2025 Oral paper.
09/2024 🎓 Began my master's journey at Tsinghua University.
08/2023 🏆 Selected as a CCF Outstanding Undergraduate, one of 99 students nationwide.

Publications

Where and When to Force: Routed Forcing for Streaming Avatars

Zihan Su, Siwen Lu, Junhao Zhuang, et al.

arXiv 2026 · First Author

Region- and noise-stage-routed distillation for more dynamic and diverse streaming avatars, while preserving visual quality and lip synchronization.

Paper Code Project Page
EchoWM: Open and Enterable Omnimodal World Models

Technical Report · Contributor

An enterable omnimodal world model with continuous navigation and synchronized video, environmental sound, music, and speech.

Paper Code Project Page
JoyAI-Echo-1.5: Long-Horizon Audio-Visual Generation for Persistent Stories and Interactive Worlds

Technical Report · Contributor

Audio-visual generation for persistent stories and interactive worlds, combining cross-shot memory, geometric camera control, and rollout-aware training.

Paper Code Project Page
Generation Enhances Understanding in Unified Multimodal Models via Multi-Representation Generation

Zihan Su, Hongyang Wei, Kangrui Cen, et al.

ICML 2026 · CCF A · First Author

Multi-representation generation for fine-grained perception, hallucination reduction, and spatial understanding.

Paper Code Project Page
Safe-Sora: Safe Text-to-Video Generation via Graphical Watermarking

Zihan Su, Xuerui Qiu, Hongbin Xu, et al.

NeurIPS 2025 · CCF A · First Author

Graphical watermarking embedded directly into video generation through hierarchical adaptive matching and 3D wavelet-enhanced Mamba.

Paper Code Project Page
TextureDiffusion: Target Prompt Disentangled Editing for Various Texture Transfer

Zihan Su, Junhao Zhuang, Chun Yuan

ICASSP 2025 Oral · CCF B · First Author

Tuning-free rich texture transfer via prompt disentanglement, structure preservation, and edit localization.

Paper Code

Education

Sep. 2024
—
Present

Tsinghua University (THU)

Computer Technology (Recommended Admission)

M.S. student, GPA 3.9/4.0.
Sep. 2020
—
Jul. 2024

Shenzhen University (SZU)

Tencent Cloud AI Class

B.Eng., Rank top 1%.

Experience

Jul. 2026
—
Present

Foundation Model Research Intern, Joy Future Academy

World Models and Real-Time Interactive Video Generation.
Apr. 2026
—
Jun. 2026

Application Intern, ByteDance Jindouyun Talent Program

Prompt Enhancement and RL-based Post-training for Text-to-Image Models.
Sep. 2025
—
Feb. 2026

Academic Research Intern, Alibaba

Unified Multimodal Generation and Understanding.
Jun. 2025
—
Sep. 2025

Application Intern, Tencent WXG

SFT and RL Post-training for Multimodal Large Language Models.
Jan. 2025
—
May 2025

Academic Research Intern, Guangming Laboratory

Diffusion Model Watermark for Video Generation.

Honors & Awards

2025Tsinghua Friend - Ordos Talent Scholarship
2024Outstanding Graduate and Honorary Bachelor's Degree, SZU.
2023CCF Outstanding Undergraduate, selected from 99 students nationwide. [Link 1] [Link 2]
2023Liyuan Star, the highest college-level honor at SZU.
2022ACM-ICPC Regional Silver Medals, Hefei and Jinan.
2022CCPC Regional Silver Medal, Guilin.
2022Ranked top 0.68% nationwide in CCF CSP.