01Multimodal reasoning
Choosing what to inspect, how long to think, and when to verify an answer.
Loading the point cloud…
Hello,I'm Kun Xiang.
01 / A LITTLE ABOUT MEGUANGZHOU, CHINA
Building the Reasoning Foundation of Physical AI
Let’s exchange ideas
I’m Kun Xiang (项鲲), a PhD candidate at Sun Yat-sen University. I work on AI that can reason about the physical world and learn by interacting with it.
I am advised by Prof. Xiaodan Liang and am currently a research intern at Alibaba DAMO Academy.
My research connects multimodal reasoning, physical world modeling, and agentic reinforcement learning. I explore how models interpret visual evidence, adapt the depth of their reasoning, and refine their decisions through feedback.
RESEARCH DIRECTIONS
02Physical world learning
Learning physical structure, dynamics, and laws from observation, simulation, and interaction.
03Agentic RL
Learning to plan, use tools, and improve decisions through reinforcement learning.
02 / RECENT UPDATES
Recent updates.
We released RynnUnify for embodied understanding, visual planning, and agentic reasoning.
One paper was accepted by TPAMI.
We released L2T and GeoThinker-F for adaptive multimodal reasoning and fast spatial intelligence.
One paper was accepted by TPAMI.
We released SeePhys Pro for diagnosing multimodal physics reasoning.
Our 3rd AI for Math Workshop was accepted by ICML 2026.
One paper was accepted by TPAMI.
We released our Physical AI survey on arXiv.
One paper was accepted by NeurIPS 2025.
We released the SeePhys code and dataset.
We released SeePhys for vision-based physics reasoning.
We released the SeePhys challenge for the 2nd AI for Math Workshop at ICML 2025.
We released AtomThink for multimodal mathematical reasoning.
03 / SELECTED RESEARCH
Research highlights.
04 / PUBLICATIONSGOOGLE SCHOLAR ↗
Publications.
Multimodal reasoning, physical intelligence, and learning systems. Newest work first. * Equal contribution or core contributor, as noted per paper.
No matching papers. Try another keyword or filter.
05 / BACKGROUND & COMMUNITY
Learning, together.
Research is a shared pursuit. Some of the places and communities that have shaped mine.
SELECTED HONORS
Outstanding Graduate of SYSU2024
National Scholarship2022
Postgraduate Scholarship of SYSU2021
RESEARCH EXPERIENCE
Alibaba DAMO Academy
CURRENTResearch Intern · May 2026 — Present
Working with Xin Li ↗
Huawei Noah’s Ark Lab
Research Intern · Jul 2024 — May 2026
Working with Jianhua Han & Hang Xu ↗
Sangfor Technologies
Research Intern · Apr — Aug 2021
Working with Jingyan Jiang ↗
ACADEMIC SERVICE
Organizer:2nd / 3rd AI for Math Workshop @ ICML ↗
Reviewer: NeurIPS, ICLR, ICML, ACL, AAAI, IEEE TNNLS, etc.
