Zhao (Bill) Xu (许钊)

Zhao (Bill) Xu (许钊)

USAIL Group

Badge GitHub followers visitors

As a new beginning, I am maintaining my new homepage – still under construction.

🐯 About Me

My Research Interest

  • My primary research focus is on building and enhancing trustworthy foundation models (e.g., LLMs, VLMs). More specifically, (i) how to better develop FMs that maintain robust performance under varying conditions while ensuring harmless and reliable responses; (ii) how to further boost the trustworthy performance of FMs without increasing model sizes or pre-training dataset sizes, particularly as the benefits from scaling laws begin to plateau. I am fortunate to have conducted some initial research in the areas of safety, robustness, and truthfulness under the guidance of all my research advisors alongside many wonderful collaborators, and I am passionate to continue studying these problems in my Ph.D.
  • I am searching for a research internship for the summer of 2025. I am available from May to September 2025. I would greatly appreciate any information or opportunities you could share. Feel free to reach out to me.

Acknowledgment

  • I used to suffer from severe lumbar and heart disease from 2020 to 2022. Despite facing severe health issues, I persevered with the support of my family, friends, and supervisors, including Prof. Yong Li and Prof. Minghua Chen, as well as Dr. Junbo Zhang, Dr. Jie Feng, Dr. Yuanwei Fang, and Kai Zhao. I am deeply grateful for their help.
  • After recovery, I am now determined and motivated to move on and make the best of each day.

Let’s Connect!

  • Feel free to reach out—whether it’s a quick question or something more in-depth! I’m always open to connecting and typically respond within 24 hours unless I’m on a tight deadline. You can reach me by email or on [Twitter/WeChat/LinkedIn—see the icons on the left]. If you’d like to schedule a quick call, you can simply click here to book a time on my Calendly. Looking forward to connecting!
Genius only means hard-working all one's life. (Mendeleev)

🔥 News

  • 2025.04: 🎉🎉 Our work, LLMLight, received the 🏆 Special Grand Award & 🏆 Golden Medal (the highest award!!) at the 50th International Exhibition of Inventions Geneva. Many thanks to Siqi for his continuous dedication and to our advisors for their huge support!

📖 Educational and Professional Background

  • Starting 2025.09, University of California, Los Angeles
    Doctor of Philosophy, Department of Computer Science
  • 2024.08 - 2024.12, The University of Chicago
    Research Intern, Department of Computer Science
    Supervisor: Prof. Bo Li
    Project: Mitigating Hallucinations in LLMs through Collaborative Multi-Agent Systems.
  • 2025 (Anticipated), The Hong Kong University of Science and Technology
    Master of Philosophy, Department of Artificial Intelligence
    Supervisor: Prof. Hao Liu Co-supervisor: Prof. Hui Xiong
    Project: Jailbreak Defense, and Evaluation Strategies for Large Language Models; Applications of Foundation Models in Smart City Scenarios.
  • 2022, Tsinghua University
    Bachelor, Department of Electronic Engineering
    Supervisor: Prof. Yong Li
  • 2019.07 - 2019.10, Pennsylvania State University
    Undergraduate Visiting Researcher, School of Electronic Engineering and Computer Science

📝 Projects and Publications

* indicates equal contribution, † indicates corresponding author.

NeurIPS 2024
sym
Arxiv 2024
sym
KDD 2024
sym
Arxiv 2024
sym

🎖 Honors and Awards

  • Special Grand Award & Golden Medal 2025 (Highest award), 50th International Exhibition of Inventions Geneva
  • Travel Grant for KDD 2024, HKUST
  • Postgraduate Scholarship (＄35,000 for stipend, ＄45,000 for HPC computing resource), HKUST
  • Scholarship for Outstanding Freshman (Full Tuition Waiver), Tsinghua University (Top 1%, 30 out of 3000)
  • Undergraduate Overseas Summer Research Scholarship, Tsinghua University
  • Outstanding Student in Work-Study Assistance, Tsinghua University

🍀 Services

  • PC Member & Reviewer: NeurIPS’25, ACL ARR’25, KDD’25, COLING’25, ICLR’25, FM-Wild/SCI-FM@ICLR’25, MARW/AIGOV@AAAI’25, SafeGenAi@NeurIPS’24, LLMAgents@ICLR’24
  • External Reviewer: NeurIPS’24, KDD’24, NeurIPS’23
  • Volunteer: KDD2024

💻 Work Experience

🎤 Invited Talks

  • 2024.11: I’ll be presenting our latest work on benchmarking LLM jailbreaks at the AI Time Forum (in Chinese). If you’re interested in the topic, please join us at 3:30 PM on November 13. I will also serve as session chair for the afternoon session on November 19.

🎁 Quick Facts about Me

Q: What motivates me to pursue research?

  • Seeking novel and engaging challenges, rather than settling for a monotonous routine.
  • Past experiences with disease, driving my desire to make impactful contributions to society.
  • Personal fulfillment in pursuing my own research interests.

Q: What type of research am I interested in?

  • Honest, intriguing, and impactful studies.
  • Research with practical, real-world applications.
  • Access to ample resources, including computational tools and expert collaborations.

Q: What are my expectations for PhD life?

  • A productive, smart, and hard-working, yet joyful experience.
  • Supportive mentoring with patience and constructive feedback.
  • Annual research internships at top-tier industry labs to connect research with real-world needs.
  • Access to adequate GPUs, to facilitate research.
  • Contribute and attend to top-tier conferences annually, striving to publish my works and broaden professional network.

As a proud member of the data mining community and an enthusiastic data miner, here are some interesting data points about me:

  • Most GPUs used for a single project: 82 A100-80G GPUs
  • Largest dataset analyzed: 10 TB
  • Collaboration experience with other researchers: 6
  • Number of top-tier research conferences attended: 2
  • Working timetable: Currently, 7 days a week, 10–14 hours per day; ideally, 6 days a week, 10–14 hours per day. I used to work very hard; however, I now believe that working smart is more important than working hard! (I received this valuable advice from several professors during my 25fall journey.)

💬 To Prospective Advisor & Collaborators

Details

I wrote this section on the last day of 2024 (Dec 31), which is a data-driven summary of my 2024 research journey.

  • Get to Know Me in 1 Minute Through Data

As I am actively seeking a Ph.D. opportunity starting in Fall 2025, I would like to share a bit more about myself—beyond my CV, SoP, and PS. With my experience in the data mining community, I believe that data provides the most faithful evidence, whether for research or life. Therefore, I would like to introduce myself in a data-driven format.

After overcoming the illness, I resumed my research journey in May 2023. Over the past 1.5 years, I have:

  1. Worked 7 days a week, 10-14 hours each day, to make up for the time lost due to illness;
  2. Led or been deeply involved in 6 research projects as the 1st or 2nd author;
  3. Released 5 research papers, with 3 accepted by NeurIPS and KDD;
  4. Open-sourced codes, datasets, and models for EVERY project (I believe in honest and reproducible research);
  5. Created or contributed to 10 GitHub academic repositories, making over 500 contributions (See my GitHub);
  6. Earned over 1.8k stars across all contributed repositories, including three repos created by me with 100+ stars—contributing 54.3% of our group’s total stars;
  7. Received 107 citations for my papers from 2023 to 2024, accounting for 79.3% of my total citations (better work, better impact);
  8. Registered for 2 conferences (though I could not attend NeurIPS due to a visa issue) and served as a student volunteer for KDD 2024;
  9. Served as a PC member or reviewer for 5 top-tier conferences/workshops, submitting 15 reviews in total;
  10. Delivered 2 talks at the KDD UFM Tutorial and AI Time Forum;
  11. Had the privilege of guidance from all my wonderful advisors and collaborated with talented friends such as Fan, Siqi, Jiawei, Dr. Jindong :), and Weijia;
  12. Obtained 3 patents for my work, which have been applied in 2 startups: Safe Intelligence AI and 阡陌 (I believe in practical research);
  13. Wrote about 28 blog posts, 51,000 words in total (writing blogs is a good way to relax and record some important moments);

I firmly believe in the power of time—the fairest and strongest force. The above details are not meant to boast about my achievements, but rather to highlight that, as an early-career researcher, I hold significant potential. Having accomplished all of this in just 1.5 years, I am confident that, given more time, I can achieve even greater results.

  • What’s Next?

In 2025, I plan to:

  1. Exercise at least 4 times a week and maintain a more balanced life and work schedule (PhD is a marathon, not a sprint);
  2. Resume my weekly reviews and planning;
  3. Strive to become 1 better person, 1 better researcher, and 1 better collaborator (PhD is a battle with yourself);
  4. Continue doing research that I want to do and that the world needs, rather than what I feel forced to do (I have followed this principle since my research journey began);
  5. Do research with more impact and practical value: deeper, broader, more important, more Github stars;
  6. Spend more time with family and friends, at least 7 hours a week (life is worth serious attention too);
  7. Read more papers—at least 1 hour per day, with 1 selected paper to read in depth to improve my writing;

Thank you for taking the time to read this section. I am sincerely grateful for every opportunity and look forward to the possibility of working with you (whether you were mentioned in my SoP or not, as our research interests and visions align).

**No matter how 2024 has been for you, I hope you have a wonderful 2025! 🌟Happy New Year!🌟**

Last update: May 4, 2025