About Me

I am Haiyang Wang (汪海洋), a researcher at Huawei (TopMinds). My research aims to build artificial superintelligence (ASI) on a foundation of coding agents for science and technology, through long-horizon agentic RL and recursive self-improvement. My broader vision is AI for Good: advancing AI to benefit people and society.

I received my Ph.D. from Peking University in 2025, advised by Prof. Liwei Wang, and my bachelor’s degree from Zhiyuan College, Shanghai Jiao Tong University, in 2020, where I worked with Prof. Cewu Lu. I have also closely collaborated with Shaoshuai Shi, with research experience at MPI-INF (with Prof. Bernt Schiele), SenseTime (with Prof. Jifeng Dai), and UCLA (with Prof. Wei Wang).

Notes: I regularly share my learning notes and reading reflections openly in AI Frontier Notes, in the hope that they are useful to others as we learn and grow together, every day.

Hiring: We recruit full-time researchers and research interns year-round, offering extensive resources and research freedom. I also welcome thoughtful discussions and informal exchanges to broaden our perspectives. Please reach out at wangocean.cs [at] gmail [dot] com.

Selected Publications

_{* means equal contribution. † indicates corresponding author.}

Preprints

• Claw-Anything: Benchmarking Always-On Personal Assistants with Broader Access to User’s Digital World.
  Yusong Lin, Xinyuan Liang, Haiyang Wang^†, Qipeng Gu, Siqi Cheng, Jiangui Chen, Shuzhe Wu, Feiyang Pan, Lue Fan, Sanyuan Zhao^†, Dandan Tu^†. [Code]    [Dataset]  

• MobileGym: A Verifiable and Highly Parallel Simulation Platform for Mobile GUI Agent Research.
  Dingbang Wu*, Rui Hao*, Haiyang Wang, Shuzhe Wu, Han Xiao, Zhenghong Li, Bojiang Zhou, Zheng Ju, Zichen Liu, Lue Fan^†, Zhaoxiang Zhang^†. [Code]  

• CLI-Gym: Scalable CLI Task Generation via Agentic Environment Inversion.
  Yusong Lin, Haiyang Wang^†, Shuzhe Wu, Lue Fan, Feiyang Pan, Sanyuan Zhao^†, Dandan Tu^†. [Code]    [Dataset]  

Publications

• FeatureBench: Benchmarking Agentic Coding for Complex Feature Development.
  Qixing Zhou*, Jiacheng Zhang*, Haiyang Wang*, Rui Hao, Jiahe Wang, Minghao Han, Yuxue Yang, Shuzhe Wu, Feiyang Pan, Lue Fan^†, Dandan Tu, Zhaoxiang Zhang^†. In ICLR 2026. [Code]    [Dataset]  

• TokenFormer: Rethinking Transformer Scaling with Tokenized Model Parameters.
  Haiyang Wang*, Yue Fan*, Muhammad Ferjad Naeem, Yongqin Xian, Jan Eric Lenssen, Liwei Wang, Federico Tombari, Bernt Schiele. In ICLR 2025. [Code]   (Spotlight Presentation, 3.20% acceptance rate)

• GiT: Towards Generalist Vision Transformer through Universal Language Interface.
  Haiyang Wang*, Hao Tang*, Li Jiang^†, Shaoshuai Shi, Muhammad Ferjad Naeem, Hongsheng Li, Bernt Schiele, Liwei Wang^†. In ECCV 2024. [Code]   (Oral Presentation, 2.32% acceptance rate)

• PRED: Pre-training via Semantic Rendering on LiDAR Point Clouds.
  Hao Yang, Haiyang Wang, Di Dai, Liwei Wang. In NeurIPS 2023.

• UniTR: A Unified and Efficient Multi-Modal Transformer for Bird’s-Eye-View Representation.
  Haiyang Wang*, Hao Tang*, Shaoshuai Shi^†, Aoxue Li, Zhenguo Li, Bernt Schiele, Liwei Wang^†. In ICCV 2023. [Code]  

• DSVT: Dynamic Sparse Voxel Transformer with Rotated Sets.
  Haiyang Wang*, Chen Shi*, Shaoshuai Shi^†, Meng Li, Sen Wang, Di He, Bernt Schiele, Liwei Wang^†. In CVPR 2023. [Code]  

• CAGroup3D: Class-Aware Grouping for 3D Object Detection on Point Clouds.
  Haiyang Wang*, Lihe Ding*, Shaocong Dong, Shaoshuai Shi^†, Aoxue Li, Jianan Li, Zhenguo Li, Liwei Wang^†. In NeurIPS 2022. [Code]  

• MsSVT: Mixed-scale Sparse Voxel Transformer for 3D Object Detection on Point Clouds.
  Shaocong Dong*, Lihe Ding*, Haiyang Wang, Tingfa Xu, Xinli Xu, Jie Wang, Ziyang Bian, Ying Wang, Jianan Li. In NeurIPS 2022. [Code]  

• RBGNet: Ray-based grouping for 3d object detection.
  Haiyang Wang, Shaoshuai Shi^†, Ze Yang, Rongyao Fang, Qi Qian, Hongsheng Li, Bernt Schiele, Liwei Wang^†. In CVPR 2022. [Code]  

• Non-convex Distributionally Robust Optimization:Non-asymptotic Analysis.
  Jikai Jin*, Bohang Zhang*, Haiyang Wang, Liwei Wang. In NeurIPS 2021.

• Explicit shape encoding for real-time instance segmentation.
  Wenqiang Xu*, Haiyang Wang*, Fubo Qi, Cewu Lu. In ICCV 2019. [Code]  

Selected Awards

• National Scholarship (博士国家奖学金), 2022-2023. Awarded annually to top 1 student in all grades of Center for Data Science, Peking University.
• National Scholarship (博士国家奖学金), 2021-2022. Awarded annually to top 1 student in all grades of Center for Data Science, Peking University.
• Excellent Student at Shanghai Jiaotong University, 2020. Top 5% of graduated students.

Experiences

• Visiting Student at MPI-INF D2. Advisor: Bernt Schiele (2023.12 - 2024.12)
• Internship at SenseTime. Advisor: Jifeng Dai and Wenguan Wang (2020.07 - 2021.07)
• Visiting Student at ScAi-UCLA. Advisor: Wei Wang (2019.09 - 2020.04)
• Undergraduate at MVIG-SJTU. Advisor: Cewu Lu (2018.02 - 2019.05)

Invited Talks

• TokenFormer: Rethinking Transformer Scaling with Tokenized Model Parameters.
  • 2024.10. Hosted by BAAI(智源). [Video]
  • 2024.11. Hosted by Huawei Noah’s Ark Lab.
  • 2024.10. Hosted by Google VIA Center
• GiT: Towards Generalist Vision Transformer through Universal Language Interface.
  • 2024.4. Hosted by Huawei Technologies Co., Ltd.
  • 2024.4. Hosted by Google VIA Center
  • 2024.1. Hosted by Prof Bernt Schiele in MPI-INF
• A Unified and Efficient Multi-Modal Transformer for Bird’s-Eye-View Representation.
  • 2023.12. Hosted by TechBeat. [Video]
  • 2023.11. Hosted by 智东西. [Video]
  • 2023.11. Hosted by 自动驾驶之心.
  • 2023.6. Hosted by Huawei Noah’s Ark Lab.

Academic Services

• Reviewer for NeurIPS’21, CVPR’22, ECCV’22, ICML’22, NeurIPS’22, CVPR’23, ICML’23, ICCV’23, NeurIPS’23, IROS’23, CVPR’24, ICML’24, NeurIPS’24 (Top Reviewer), ICLR’25, ICLR’26