Hi, I am Wentao Hu, a Master's student in the College of Computing and Data Science (CCDS) at Nanyang Technological University (NTU), advised by Prof. Hanwang Zhang in Mreal Lab. Before that, I earned my Bachelor's degree in Statistics from Hunan University(HNU).
My research interests lie in AIGC and MLLM, and I am eager to explore the unknown.
Warning
Problem: The current name of your GitHub Pages repository ("Solution: Please consider renaming the repository to "
http://".
However, if the current repository name is intended, you can ignore this message by removing "{% include widgets/debug_repo_name.html %}" in index.html.
Action required
Problem: The current root path of this site is "baseurl ("_config.yml.
Solution: Please set the
baseurl in _config.yml to "Education
-
Nanyang Technological UniversityCollege of Computing and Data Science
M.Eng. StudentAug. 2023 - present -
Hunan UniversityB.S. in StatisticsSep. 2019 - Jun. 2023
Experience
-
Central Media Technology Institute(Singapore), Huawei 2012 laboratoryAI Research InternAug. 2024 - May. 2025 -
Kuaishou TechnologyInternDec. 2022 - Feb. 2023
News
Publications
On Path to Multimodal Generalist: Levels and Benchmarks
Hao Fei, Yuan Zhou, Juncheng Li, Xiangtai Li, Qingshan Xu, Bobo Li, Shengqiong Wu, Yaoting Wang, Junbao Zhou, Jiahao Meng, Qingyu Shi, Zhiyuan Zhou, Liangtao Shi, Minghe Gao, Daoan Zhang, Zhiqi Ge, Siliang Tang, Kaihang Pan, Yaobo Ye, Haobo Yuan, Tao Zhang, Weiming Wu, Tianjie Ju, Zixiang Meng, Shilin Xu, Liyu Jia, Wentao Hu, Meng Luo, Jiebo Luo, Tat-Seng Chua, Hanwang Zhang, Shuicheng Yan
International Conference on Machine Learning (ICML) 2025 (Oral),
We propose a General-Level framework, inspired by the five-level capability grading mechanisms in the autonomous driving industry, to assess the performance and generality of Multimodal Language Models (MLLMs) across five levels. Central to this framework is the concept of Synergy, which categorizes capabilities based on whether MLLMs preserve synergy across comprehension, generation, and multimodal interactions. To evaluate the comprehensive abilities of various generalists, we present General-Bench, a massive, ever-growing multimodal benchmark that encompasses a broad spectrum of skills, modalities, formats, and capabilities — with over 700 tasks and 325,800 instances.
[arXiv] [Project Page] [Leaderboard] [Huggingface Benchmark] [新智元] [机器之心]
On Path to Multimodal Generalist: Levels and Benchmarks
Hao Fei, Yuan Zhou, Juncheng Li, Xiangtai Li, Qingshan Xu, Bobo Li, Shengqiong Wu, Yaoting Wang, Junbao Zhou, Jiahao Meng, Qingyu Shi, Zhiyuan Zhou, Liangtao Shi, Minghe Gao, Daoan Zhang, Zhiqi Ge, Siliang Tang, Kaihang Pan, Yaobo Ye, Haobo Yuan, Tao Zhang, Weiming Wu, Tianjie Ju, Zixiang Meng, Shilin Xu, Liyu Jia, Wentao Hu, Meng Luo, Jiebo Luo, Tat-Seng Chua, Hanwang Zhang, Shuicheng Yan
International Conference on Machine Learning (ICML) 2025 (Oral),
We propose a General-Level framework, inspired by the five-level capability grading mechanisms in the autonomous driving industry, to assess the performance and generality of Multimodal Language Models (MLLMs) across five levels. Central to this framework is the concept of Synergy, which categorizes capabilities based on whether MLLMs preserve synergy across comprehension, generation, and multimodal interactions. To evaluate the comprehensive abilities of various generalists, we present General-Bench, a massive, ever-growing multimodal benchmark that encompasses a broad spectrum of skills, modalities, formats, and capabilities — with over 700 tasks and 325,800 instances.
[arXiv] [Project Page] [Leaderboard] [Huggingface Benchmark] [新智元] [机器之心]
Selftok: Discrete Visual Tokens of Autoregression, by Diffusion, and for Reasoning
Bohan Wang, Zhongqi Yue, Fengda Zhang, Shuo Chen, Li'an Bi, Junzhe Zhang, Kennard Yanting Chan, Jiachun Pan, Weijia Wu, Mingze Zhou, Wang Lin, Kaihang Pan, Saining Zhang, Liyu Jia, Wentao Hu, Wei Zhao, Hanwang Zhang
arXiv Preprint,Under Review 2025
We completely discard the conventional spatial prior in image representation and introduce a novel discrete visual tokenizer: Self-Consistency Tokenizer (Selftok). Selftok is a SOTA tokenizer that achieves both high-quality reconstruction and high compression bit rate. After representing the training images as Selftok tokens, as a pure AR model, our VLM achieves both SOTA visual comprehension and generation performances.
Selftok: Discrete Visual Tokens of Autoregression, by Diffusion, and for Reasoning
Bohan Wang, Zhongqi Yue, Fengda Zhang, Shuo Chen, Li'an Bi, Junzhe Zhang, Kennard Yanting Chan, Jiachun Pan, Weijia Wu, Mingze Zhou, Wang Lin, Kaihang Pan, Saining Zhang, Liyu Jia, Wentao Hu, Wei Zhao, Hanwang Zhang
arXiv Preprint,Under Review 2025
We completely discard the conventional spatial prior in image representation and introduce a novel discrete visual tokenizer: Self-Consistency Tokenizer (Selftok). Selftok is a SOTA tokenizer that achieves both high-quality reconstruction and high compression bit rate. After representing the training images as Selftok tokens, as a pure AR model, our VLM achieves both SOTA visual comprehension and generation performances.
Reasoning Physical Video Generation with Diffusion Timestep Tokens via Reinforcement Learning
Wang Lin*, Liyu Jia*, Wentao Hu*, Kaihang Pan, Zhongqi Yue, Jingyuan Chen, Fei Wu, Hanwang Zhang (* co-first authors)
arXiv Preprint,Under Review 2025
We propose an autoregressive-based video generation framework Phys-AR that incorporates a symbolic reasoning process into the generation process, thus maintaining the physical correctness of the generated videos.
Reasoning Physical Video Generation with Diffusion Timestep Tokens via Reinforcement Learning
Wang Lin*, Liyu Jia*, Wentao Hu*, Kaihang Pan, Zhongqi Yue, Jingyuan Chen, Fei Wu, Hanwang Zhang (* co-first authors)
arXiv Preprint,Under Review 2025
We propose an autoregressive-based video generation framework Phys-AR that incorporates a symbolic reasoning process into the generation process, thus maintaining the physical correctness of the generated videos.