👋 About Me
I am currently a PhD student in Data Science and Information Technology at Tsinghua University, devoted to advancing general-purpose agents that can perceive, reason, and act. My main focus is on scientific reasoning, world models, and multi-agent systems. Prior to this, I earned my B.Eng. in Communication Engineering from East China Normal University.
Since May 2026, I have been a Research Intern at Alibaba ATH Token Foundry, working on video world models. Previously, I was a Research Intern at the PRIME-RL Team of Shanghai AI Laboratory, working with Ganqu Cui and Prof. Ning Ding on post-training of foundation models. Before that, I was a visiting student at the Qing Yuan Research Institute, Shanghai Jiao Tong University, working with Prof. Guohao Dai. I also maintain long-term collaborations with Junchi Yu and Prof. Philip Torr at the University of Oxford.
My recent work spans scientific reasoning, world models and embodied AI, multi-agent systems, and efficient / continual LLMs.
I am actively looking for research collaborations and discussions. Feel free to reach out! (wanhy24@mails.tsinghua.edu.cn).
🔥 News
- 2026.06: 🎉 One paper accepted to the 3rd AI for Math Workshop: Toward Self-Evolving Scientific Agents at ICML 2026 (P1-VL).
- 2026.04: 🎉 One paper accepted to IJCAI 2026 (SDFLoRA).
- 2026.04: 🎉 Two papers accepted to ICML 2026 (HiPhO, LabBuilder).
- 2026.04: 🎉 One paper accepted to ACL 2026 (Multi-Agent MCTS for LLM Inductive Reasoning).
- 2026.03: 🎉 Autonomous Laboratory Agent demo released — an embodied Vision-Language-Action agent that autonomously performs wet-lab chemistry experiments.
- 2026.02: 🎉 P1-VL-235B-A22B released — extending P1 to multimodal physics reasoning with visual perception.
- 2026.01: 🎉 One paper accepted to ICLR 2026 (From What to Why).
- 2025.12: 📰 From Tokens to Frames highlighted by 量子位.
- 2025.11: 📰 P1-235B-A22B highlighted by 量子位.
- 2025.11: 🎉 P1-235B-A22B released — the first open-source physics reasoning model trained purely via reinforcement learning to attain gold-medal performance on IPhO 2025, sweeping 12 Gold + 1 Silver across 13 international and regional physics olympiads and rivaling frontier closed-source systems such as GPT-5 and Gemini-2.5-Pro.
- 2025.11: 🎉 One paper accepted to AAAI 2026 (DeepResearch Arena).
- 2025.09: 🎉 One paper accepted to NeurIPS 2025 (Spotlight Attention).
- 2025.08: 🎉 One paper accepted to EMNLP 2025 (RECALL).
📖 Education
Sep. 2024 – Jul. 2029 (expected): Ph.D. in Data Science and Information Technology, Tsinghua University (THU), Beijing, China.
Sep. 2020 – Jul. 2024: B.Eng. in Communication Engineering, East China Normal University (ECNU), Shanghai, China.
📑 Technical Report
P1: Mastering Physics Olympiads with Reinforcement Learning.
Shanghai AI Lab PRIME-RL Team · Haiyuan Wan (Core Contributor)
Technical Report, 2026.
Project ·
Paper ·
Code ·
P1-235B-A22B ·
P1-30B-A3B ·
量子位
P1-VL: Bridging Visual Perception and Scientific Reasoning in Physics Olympiads.
Shanghai AI Lab PRIME-RL Team · Haiyuan Wan
Technical Report, 2026.
Project ·
Paper ·
Code ·
P1-VL-235B-A22B ·
P1-VL-30B-A3B
📝 Selected Publications
DeepResearch Arena: The First Exam of LLMs’ Research Abilities via Seminar-Grounded Tasks.
Haiyuan Wan†, Chen Yang†, Junchi Yu, Meiqi Tu, Jiaxuan Lu, Di Yu, Jianbao Cao, Ben Gao, Jiaqing Xie, Aoran Wang, Wenlong Zhang, Philip Torr, Dongzhan Zhou.
(† equal contribution) AAAI Conference on Artificial Intelligence (AAAI), 2026.
HiPhO: How Far Are (M)LLMs from Humans in the Latest High School Physics Olympiad Benchmark?
Fangchen Yu†, Haiyuan Wan†, Qianjia Cheng†, Yuchen Zhang, Jiacheng Chen, Fujun Han, Yulun Wu, Junchi Yao, Ruilizhen Hu, Ning Ding, Yu Cheng, Tao Chen, Lei Bai, Dongzhan Zhou, Yun Luo, Ganqu Cui, Peng Ye.
(† equal contribution) International Conference on Machine Learning (ICML), 2026.
RECALL: REpresentation-aligned Catastrophic-forgetting ALLeviation via Hierarchical Model Merging.
Bowen Wang†, Haiyuan Wan†, Liwen Shi, Chen Yang, Peng He, Yue Ma, Haochen Han, Wenhao Li, Tiao Tan, Yongjian Li, Fangming Liu, Yifan Gong, Sheng Zhang.
(† equal contribution) Conference on Empirical Methods in Natural Language Processing (EMNLP), 2025.
From What to Why: A Multi-Agent System for Evidence-based Chemical Reaction Condition Reasoning.
Cheng Yang, Jiaxuan Lu, Haiyuan Wan, Junchi Yu, Feiwei Qin.
International Conference on Learning Representations (ICLR), 2026.
From Tokens to Frames: Video Generation as a New Paradigm for Spatial Reasoning.
Cheng Yang†, Haiyuan Wan†, Yiran Peng†, Xin Cheng, Zhaoyang Yu, Jiayi Zhang, Junchi Yu, Xinlei Yu, Xiawu Zheng, Dongzhan Zhou, Chenglin Wu.
(† equal contribution) arXiv, 2025.
Spotlight Attention: Towards Efficient LLM Generation via Non-linear Hashing-based KV Cache Retrieval.
Wenhao Li, Yuxin Zhang, Gen Luo, Haiyuan Wan, Ziyang Gong, Fei Chao, Rongrong Ji.
Conference on Neural Information Processing Systems (NeurIPS), 2025.
PhysicsMinions: Winning Gold Medals in the Latest Physics Olympiads with a Coevolutionary Multimodal Multi-Agent System.
Fangchen Yu†, Junchi Yao†, Ziyi Wang, Haiyuan Wan, Youling Huang, Bo Zhang, Shuyue Hu, Dongzhan Zhou, Ning Ding, Ganqu Cui, Lei Bai, Wanli Ouyang, Peng Ye.
(† equal contribution) arXiv, 2025.
LabBuilder: Protocol-Grounded 3D Layout Generation for Interactable and Safe Laboratory.
Jianbao Cao†, Zhangrui Zhao†, Bohan Feng†, Zixuan Hu, Rui Li, Haiyuan Wan, Chenxi Li, Jingyuan Li, Wenzhe Cai, Lei Bai, Wanli Ouyang, Lingyu Duan, Di Huang, Mingting Pan, Sha Zhang, Xinzhu Ma, Shixiang Tang, Dongzhan Zhou.
(† equal contribution) International Conference on Machine Learning (ICML), 2026.
🎬 Demos
Autonomous Laboratory Agent
Shanghai AI Lab Physical Intelligence Center · Haiyuan Wan (Core Contributor)
Demo, 2026.
Demo💼 Internships
- May 2025 – May 2026: Shanghai AI Laboratory · Research Intern · PRIME-RL Team · Post-training of foundation models
- May 2026 – Present: Alibaba ATH · Research Intern · Token Foundry · Video world models
🎖 Honors and Awards
- Junhao Foundation Soaring Scholarship — Presented by Academician Chu Junhao; the only undergraduate awardee.
- ECNU Outstanding Student Special Scholarship — Sole recipient in the department, awarded for two consecutive years.
💬 Services
- Reviewer: ICML, NeurIPS, ICLR, AAAI, ECCV and other top-tier conferences/journals in computer vision, natural language processing, and machine learning.