Yue Su 苏悦

PhD Student · MMLab, The University of Hong Kong

Education

The University of Hong KongSep 2026 – Present PhD, MMLab · Department of Electrical and Computer Engineering Advisor: Prof. Xihui Liu
Xidian UniversitySep 2022 – Jun 2026 B.Eng., School of Artificial Intelligence Rank: 4/174 · GPA: 3.7/4.0 · National Scholarship

Experience

Tencent IEGLightSpeed StudiosSep 2026 – Present Talent Program Intern
ByteDance SeedSeed RoboticsOct 2025 – Apr 2026 Research Intern
Astribot Inc.Research DepartmentJun 2025 – Sep 2025 Research Internwith Jianan Wang
Shanghai Jiao Tong UniversityMVIGJul 2024 – Jun 2025 Research Assistantwith Prof. Cewu Lu and Prof. Lixin Yang
Xidian UniversityCooperative Intelligent Systems LabSep 2023 – Jul 2024 Internwith Prof. Maoguo Gong and Prof. Hao Li

Work

  • Modular Connections: The Connections Between Residual Streams Can Themselves Be Modular Models Yue Su Blog 2026English中文
  • NextTransformer: An Open Framework and Fair Benchmark for Next-Generation Transformer Architectures Compute-matched training and audited results across Transformer families. Initiator of this project Project 2026WebsiteDocsCode
  • OneWM: Can World and Action Modeling Benefit Each Other under Unified Causal Training? Yue Su*, Bokai Lin*, Xinkai Wang*, Chubin Zhang, Xiangnan Wu, Kuan Zhang, Weiyang Jin, Jiwen Yu, Yiming Li, Cewu Lu, Lixin Yang†, Xihui Liu† Preprint 2026arXivCodeProjectVideo
  • World Guidance: World Modeling in Condition Space for Action Generation Yue Su, Sijin Chen, Haixin Shi, Mingyu Liu, Zhengshen Zhang, Ningyuan Huang, Weiheng Zhong, Zhengbang Zhu, Yuxiao Liu†, Xihui Liu† ICML 2026arXivCodeProject
  • DSPv2: Improved Dense Policy for Effective and Generalizable Whole-body Mobile Manipulation Yue Su, Chubin Zhang, Sijin Chen, Liufan Tan, Yansong Tang, Jianan Wang, Xihui Liu† ICRA 2026arXivCodeProject
  • Dense Policy: Bidirectional Autoregressive Learning of Actions Yue Su*, Xinyu Zhan*, Hongjie Fang, Han Xue, Haoshu Fang, Yong-Lu Li, Cewu Lu, Lixin Yang† ICCV 2025PaperarXivCodeProject
1 / 2
  • Motion Before Action: Diffusing Object Motion as Manipulation Condition Yue Su*, Xinyu Zhan*, Hongjie Fang, Yong-Lu Li, Cewu Lu, Lixin Yang† RA-L 2025, ICRA 2026PaperarXivCodeProject
  • EgoForesight: Effectively Scaling Policy Learning with Noisy Egocentric Actions Chubin Zhang, Changyuan Wang, Yirui Wang, Ao Li, Shilin Ma, Yue Su, Cai Zhou, Xiuwei Xu, Tianle Zhang, Yansong Tang† Preprint 2026arXivCodeProject
  • Truncate or Not? A Scaling Study of Looped Transformers with Truncated Backpropagation Xuyuan Xiong*, Siyu Chen*, Zihan Shao, Yue Su, Weiyang Jin, Ping Luo, Zhuoran Yang†, Xihui Liu† Preprint 2026arXivCodeProject
  • Scaling in Games: Continued Pre-Training for Embodied Agents in Diverse Virtual Worlds Kuan Zhang*, Yukun Chen*, Zhihao Yang*, Yue Su, Xiangnan Wu, Zirong Chen, Run Luo, Jinkun Hou, Tao Tan, Yinhe Zheng†, Yiming Li† Preprint 2026PaperCodeProject
  • Video-Arena: A Scalable Agentic Video Benchmark for Spatiotemporal Capability Kuan Zhang*, Yukun Chen*, Zhihao Yang*, Yue Su, Xiangnan Wu, Zirong Chen, Run Luo, Tao Tan, Yinhe Zheng†, Yiming Li† Preprint 2026arXivCodeProject
  • SG-WAM: Self-Guided World Modeling in Geometry-Aware Policy Space Ruiteng Zhao, Zhengshen Zhang, Yue Su, Wenshuo Wang, Jiahui Li, Zhiyuan Yang, Francis E. H. Tay, Marcelo H. Ang Jr., Haiyue Zhu† Preprint 2026arXivCodeProject
  • CARA: Concept-Aware Risk Attention for Interpretable Collision Prediction Zhishan Tao, Ruoyu Wang, Yucheng Wu, Enjun Du, Yilei Yuan, Sherwin Ho, Yue Su, Jinbo Su, Yi Hong† ACM MM 2026OralarXiv
  • Towards Generalist Game Players: An Investigation of Foundation Models in the Game Multiverse Kuan Zhang*, Dongchen Liu*, Qiyue Zhao*, Tianyu Xin*, Yue Su*, Haisheng Wang, …, Yiming Li† TMLR 2026arXivCode
  • CLAP: Contrastive Latent Action Pretraining for Learning Vision-Language-Action Models from Human Videos Chubin Zhang*, Jianan Wang*, Zifeng Gao, Yue Su, Tiranru Dai, Cai Zhou, Jiwen Lu, Yansong Tang† TRO 2026arXivCodeProject
  • AdvDisplay: Adversarial Display Assembled by Thermoelectric Cooler for Fooling Thermal Infrared Detectors Hao Li†, Fanggao Wan, Yue Su, Yue Wu, Mingyang Zhang, Maoguo Gong† AAAI 2025Paper
  • Aha Looped Transformer Interactive atlas of looped Transformers, adaptive depth, and latent reasoning. Project 2026WebsiteCode
  • ManiUniCon: A Unified Control Interface for Robotic Manipulation Unified robot control with sensor integration and real-time policy execution. Project 2025TeamCode

Invited Talks

Awards

  • Xiaomi Outstanding Scholarship
  • National Scholarship
  • Outstanding Student, Xidian University

Leadership & Community

  • Founder, IAAI
  • Head of the Research Department, Microsoft Club, Xidian University