I am a third-year Ph.D. candidate in the
Department of Computer Science and Technology,
Tsinghua University,
advised by Prof.
Jun Zhu.
Before that, I received my B.E. degree from the Department of Computer Science and Technology, Tsinghua University,
and my secondary B.S. degree from the
Department of Psychology,
Tsinghua University, in June, 2022.
I work closely with Prof.
Hang Su and Prof.
Yinpeng Dong. I am also honored to have visited TML Lab led by Prof. Tongliang Liu.
My research mainly focuses on machine learning, deep learning and their applications in computer vision, etc.
Recently, I am interested in the robustness, safety and trustworthiness of deep learning models, especially general-purpose large models.
I look forward to collaborating with leading research groups in trustworthy AI and related fields.
I am open to opportunities for collaboration and academic exchange in any form. Please feel free to contact me if you are interested.
News
- [2025-05] We are oranizing a Workshop on Safe and Trustworthy Multimodal AI Systems at ICCV 2025! Find out more at https://safemmai.github.io
- [2025-05] Our paper on safe alignment with introspective reasoning (STAIR) is accepted as a Spotlight at ICML 2025!
This work pioneers improving safety alignment with introspective reasoning. After three iterations of self-improvement and with test-time search, we boost the goodness score of Llama-3.1-8B-Instruct on StrongReject to 0.94, comparative to Claude3.5-Sonnet!
- [2025-04] We release the first version of RealSafe-R1, a safety-aligned DeepSeek-R1 family without compromising reasoning capability.
- [2024-12] We are oranizing a Workshop on Test-time Scaling for Computer Vision at CVPR 2025! Find out more at https://viscale.github.io
- [2024-09] Two benchmark papers accepted at NeurIPS 2024!
MultiTrust focuses on the trustworthiness of Multimodal LLMs, testing more than 20 modern MLLMs on 32 carefully curated tasks. Find out more at https://multi-trust.github.io
- [2024-03] One paper accepted as a Highlight at CVPR 2024!
- [2024-01] One paper accepted at ICLR 2024!
- [2023-03] One paper accepted at CVPR 2023!
- [2023-03] One paper accepted in CVIU!
- [2022-12] Our team (Yinpeng Dong, Chang Liu, Wenzhao Xiang, Yichi Zhang, Haoxing Ye) won the 1st place in the Evaluating the Adversarial Robustness of Deep Learning Model track in 2022 International Algorithm Case Competition.
- [2022-06] I received Beijing Outstanding Graduates (Only 4 students in the department of CS).
Selected Publications
Full publications can be found at Google Scholar.
* for equal contribution
Education
- Ph.D. in Computer Science, Tsinghua University, 2022-2027
- B.E. in Computer Science, Tsinghua University, 2018-2022
- B.S. (Secondary) in Psychology, Tsinghua University, 2019-2022
Experiences
- Research Intern, Qwen Team, Alibaba, 2025-Present
- Visiting Ph.D. Student, TML Lab led by Prof. Tongliang Liu, University of Sydney, Spring, 2025
- Research Intern, RealAI, 2023-2025
- Research Intern, Tencent, 2021
Awards & Honors
- Spotlight Poster, ICML 2025
- Highlight Poster, CVPR 2024
- 1st place in the Evaluating the Adversarial Robustness of Deep Learning Model track in 2022 International Algorithm Case Competition, 2023
- Outstanding Graduate, Tsinghua University, Beijing, 2022
- Merit Student, Beijing, 2021
- 2nd place (2/1599) in the CVPR 2021 Security AI Challenger Unrestricted Adversarial Attacks on ImageNet, 2021
- 8th place (8/1814) in the CIKM 2020 Challenge Adversarial Challenge on Object Detection, 2020
- Comprehensive Excellence Scholarships, Tsinghua University, 2019-2022
- 8th top scorer of science in NCEE (Gaokao), Beijing, 2018