PeopleofSuperintelligence

Researcher

Yinpeng Dong

Yinpeng Dong is an AI researcher whose work examines model robustness and trustworthiness. He earned computer science degrees at Tsinghua and coauthored work on adversarial attacks and robustness benchmarks. His later research includes diffusion-based classification, multimodal model trust, and reasoning-based language-model safety alignment. As of October 2026, he is an assistant professor and assistant to the dean at Tsinghua's College of AI and the principal investigator of T-STAR Lab.

Timeline

  1. 2017–2022

    Tsinghua computer science degrees

    At Tsinghua's Department of Computer Science and Technology, he earns a B.E. in 2017 and a Ph.D. in 2022, advised by Professor Jun Zhu.

  2. 2018

    Momentum-based adversarial attacks

    At CVPR in 2018, he coauthors a momentum-based iterative method that stabilizes attack updates and produces more transferable adversarial examples.

  3. April 2019

    A VALSE student paper award

    In April 2019, the VALSE Annual Outstanding Student Paper Award goes to “Boosting Adversarial Attacks with Momentum,” his 2018 CVPR paper.

  4. 2020

    A benchmark for robustness

    In 2020, he coauthors a benchmark for adversarial robustness in image classification; its large-scale experiments use two robustness curves to evaluate attack and defense methods.

  5. January 2022–February 2025

    Postdoctoral research at Tsinghua

    From January 2022 to February 2025, he works as a postdoctoral researcher in Tsinghua's Department of Computer Science and Technology.

  6. December 2022

    A CCF dissertation award

    In December 2022, his doctoral thesis on adversarial attacks and robustness evaluation in deep learning is selected for the CCF Outstanding Doctoral Dissertation Award.

  7. 2024

    A robust diffusion classifier

    With Huanran Chen and colleagues, he introduces Robust Diffusion Classifier in a 2024 ICML paper, using a pretrained diffusion model as a generative classifier for adversarial robustness.

  8. 2024

    MultiTrust tests multimodal models

    With collaborators, he coauthors MultiTrust in 2024, a benchmark that evaluates multimodal large language models for truthfulness, safety, robustness, fairness, and privacy.

  9. 2025

    STAIR for safety alignment

    At ICML in 2025, he coauthors STAIR, which uses safety-aware step-by-step reasoning to help language models identify risks.

  10. Since February 2025

    Assistant professor at Tsinghua

    In February 2025, he becomes an Assistant Professor at Tsinghua's College of AI. As of October 2026, he is also Assistant to the Dean and PI of T-STAR Lab, whose research spans AI theory, machine learning, large-model safety and alignment, and generative AI.

Career

  1. Since 2025Tsinghua UniversityAssistant Professor

Education

  1. 2022Tsinghua UniversityPh.D.
  2. 2017Tsinghua UniversityB.E.

From the biography.

Checked career sources: collegeai.tsinghua.edu.cn, dongyp13.github.io.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Yinpeng Dong yet. Send us a free-licensed photo.

  1. Yinpeng Dong dongyp13.github.io. Checked 26 October 2026.
  2. Yinpeng DONG - College of AI, Tsinghua University collegeai.tsinghua.edu.cn. Checked 26 October 2026.
  3. Boosting Adversarial Attacks With Momentum openaccess.thecvf.com. Checked 26 October 2026.
  4. Benchmarking Adversarial Robustness on Image Classification openaccess.thecvf.com. Checked 26 October 2026.
  5. Robust Classification via a Single Diffusion Model proceedings.mlr.press. Checked 26 October 2026.
  6. MultiTrust: A Comprehensive Benchmark Towards Trustworthy Multimodal Large Language Models proceedings.neurips.cc. Checked 26 October 2026.
  7. STAIR: Improving Safety Alignment with Introspective Reasoning proceedings.mlr.press. Checked 26 October 2026.
  8. Research Groups - College of AI, Tsinghua University collegeai.tsinghua.edu.cn. Checked 26 October 2026.