Yukai Zhou (周宇凯)

---

Biography

I am currently a second year CS Ph.D. student at ShanghaiTech University, to be fortunately advised by Prof. Wenjie Wang. Prior to that, I received my Bachelor's degree in Physics from ShanghaiTech University in 2023, and pursued the CS Master's degree from 2023 to 2025. My research interest lies in real-world AI safety issues as well as their mitigation techniques, with a specific emphasis on language model behavior control and adversarial manipulations.

---

News

• [Oct. 2026] Invited talk at the 1-st "Jiao-Chuang" Global PhD Student Forum, Shanghai Jiao Tong University.
• [Jun. 2026] One patent application has been filed.
• [Jun. 2025] My first work DSN is accepted to ACL 2025 Findings. 🎉🎉🎉
• [Apr. 2025] Invited as a guest lecturer for CS246: Trustworthy ML course (LLM Jailbreaking).
• [Oct. 2024] Secure the first place in JailbreakBench, an open-sourced jailbreak leaderboard.
• [Oct. 2024] Secure the best white-box method in CLAS, a NeurIPS 2024 Contest.
• [Aug. 2024] Awarded "Outstanding Student" (top 10%) for 2023–2024 academic year.
• [Sep. 2023] Joined ASPIRE Lab, to begin my exploration in CS, eventually.

---

Publications & Preprints

• Rethinking LLM Jailbreak Evaluation: Continuous Risk Landscapes Beyond Binary Outcomes
  Yukai Zhou, Wenjie Wang
  arXiv 2026. [PDF] [Code] [Website]

• How Jailbreak Attacks Inform Safety Alignment: A Defender-Centric, Shapley-Based Evaluation of Jailbreak Contributions
  Yukai Zhou, Feiyang Lu, Xiaokai Mao, Jinfei Liu, Wenjie Wang
  arXiv 2026. [PDF] [Code] [Website] [Chinese Blog]

• Beyond Jailbreaks: Revealing Stealthier and Broader LLM Security Risks Stemming from Alignment Failures
  Yukai Zhou, Sibei Yang, Wenjie Wang
  arXiv 2025. [PDF] [Code] [Website] [Dataset]

• Don’t Say No: Jailbreaking LLM by Suppressing Refusal
  Yukai Zhou, Jian Lou, Zhijie Huang, Zhan Qin, Sibei Yang, Wenjie Wang
  ACL 2025 Findings. [PDF] [Code] [Blog in Chinese] [Leaderboard of JailbreakBench]

---

Research Interests

• Real-world AI Safety Issues
  - Jailbreak Attacks on Large Language Models.
  - Real-world Consequences of Jailbreaks.
  - Agent Safety.

• AI Safety Mitigation Methods
  - Alignment & LLM Post-training.
  - Adversarial Attacks & Training.
  - Defensive Techniques.

---

Awards and Services

• Reviewer: ACL Rolling Review (2025, 2026), ACM Computing Surveys, ICLR 2027.
  

• Guest Lecturer: Lectures on LLM jailbreaking for CS246: Trustworthy ML (2025, 2026).
  

• Outstanding Student: Awarded to the top 10%
  

• Teaching Assistant: Introduction to Information Science and Technology, SI100b, Spring 2024.
  

• Undergraduate Mentor: Dadao College, Sep. 2023 – Jan. 2024.