Kun Xiang

Shenzhen, China

Thanks for stopping by! 👋

I am currently a PhD student at the HCP-I2 Lab in Sun Yat-sen University advised by Prof. Xiaodan Liang. Grounded in data-centric perspective, our work aims to build generalizable multimodal AI systems. Previously, I obtained both a Bachelor’s and Master’s degree at SYSU under the supervision of Prof. Shancheng Jiang, concentrating on Trustworthy Machine Learning and Computer Aided Diagnosis System.

Currently, I am interested in the higher-order reasoning abilities of MLLMs. I am working to answer: 1) How can we search for optimal thinking strategies? 2) How to evaluate the reasoning ability of LLMs/MLLMs?

Feel free to reach out at xiangk123@gmail.com to discuss potential collaboration.

news

May 21, 2025	🔥 We release the paper of SeePhys. It is a full spectrum multimodal benchmark for evaluating physics reasoning across different knowledge levels!
May 21, 2025	🎉 SeePhys is officially open for challenges at the 2nd AI for Math Workshop at ICML 2025!
Mar 8, 2025	We announce AtomThink, a novel slow thinking framework for multimodal mathematical reasoning task!
Jul 1, 2024	One paper accepted by Information Sciences, IF=8.1.
Jun 26, 2023	One paper accepted by AAAI 2024! See you in Washington!

selected publications

SeePhys: Does Seeing Help Thinking? – Benchmarking Vision-Based Physics Reasoning

Kun Xiang, Heng Li, Terry Jingchen Zhang, Yinya Huang, and 10 more authors

arxiv.org/pdf/2505.19099v1, 2025

Bib HTML Code

@article{xiang2025seephys,
  title = {SeePhys: Does Seeing Help Thinking? -- Benchmarking Vision-Based Physics Reasoning},
  author = {Xiang, Kun and Li, Heng and Zhang, Terry Jingchen and Huang, Yinya and Liu, Zirong and Qu, Peixin and He, Jixi and Chen, Jiaqi and Yu-Jie, Yuan and Han, Jianhua and Xu, Hang and Li, Hanhui and Sachan, Mrinmaya and Liang, Xiaodan},
  journal = {arxiv.org/pdf/2505.19099v1},
  year = {2025},
  huggingface = {https://huggingface.co/datasets/SeePhys/SeePhys},
}

Can Atomic Step Decomposition Enhance the Self-structured Reasoning of Multimodal Large Models?

Kun Xiang, Zhili Liu, Zihao Jiang, Yunshuang Nie, and 7 more authors

arXiv preprint arXiv:2503.06252, 2025

Bib HTML PDF Code

@article{xiang2025can,
  title = {Can Atomic Step Decomposition Enhance the Self-structured Reasoning of Multimodal Large Models?},
  author = {Xiang, Kun and Liu, Zhili and Jiang, Zihao and Nie, Yunshuang and Cai, Kaixin and Yin, Yiyang and Huang, Runhui and Fan, Haoxiang and Li, Hanhui and Huang, Weiran and others},
  journal = {arXiv preprint arXiv:2503.06252},
  year = {2025},
}

Toward robust diagnosis: A contour attention preserving adversarial defense for covid-19 detection

Kun Xiang, Xing Zhang, Jinwen She, Jinpeng Liu, and 3 more authors

In Proceedings of the AAAI Conference on Artificial Intelligence, 2023

Bib HTML Code

@inproceedings{xiang2023toward,
  title = {Toward robust diagnosis: A contour attention preserving adversarial defense for covid-19 detection},
  author = {Xiang, Kun and Zhang, Xing and She, Jinwen and Liu, Jinpeng and Wang, Haohan and Deng, Shiqi and Jiang, Shancheng},
  booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},
  volume = {37},
  number = {3},
  pages = {2928--2937},
  year = {2023},
}

A novel weight pruning strategy for light weight neural networks with application to the diagnosis of skin disease

Kun Xiang, Linlin Peng, Haiqiong Yang, Mingxin Li, and 3 more authors

Applied Soft Computing, 2021

Bib PDF

@article{xiang2021novel,
  title = {A novel weight pruning strategy for light weight neural networks with application to the diagnosis of skin disease},
  author = {Xiang, Kun and Peng, Linlin and Yang, Haiqiong and Li, Mingxin and Cao, Zhongfa and Jiang, Shancheng and Qu, Gang},
  journal = {Applied Soft Computing},
  volume = {111},
  pages = {107707},
  year = {2021},
  publisher = {Elsevier},
}