Yihua Zhang

Ph.D. student in CS at @michiganstateu. Trustworthy and Scalable AI. IBM PhD Fellowship Awardee 2024, Previously @Amazon, @Cisco, and currently @MetaAI.

[Yihua Zhang @zyh2022]

Interesting results! Great job, Junyuan!

[Junyuan "Jason" Hong @hjy836]

New Finding: 🧠 LLMs Can Get Brain Rot (too)!

[1/6]🚨LLMs can suffer from “brain rot” when continually fed trivial, highly engaging Twitter/X content.

🧩Their reasoning, long‑context understanding, safety, and even personality traits persistently deteriorate.

[Bolei Zhou @zhoubolei]

@NeurIPSConf This is great! But will you also consider setting up an official satellite location in China, given the fact that so many great NeurIPS papers come from China and so many Chinese researchers couldn't attend the conference due to the US/Canada Visa issue?

[Chongyu Fan @ChongyuFan]

Excited to share our ICML '25 work on robust LLM unlearning—“Towards LLM Unlearning Resilient to Relearning Attacks: A Sharpness-Aware Minimization Perspective and Beyond”.

🔗 Paper: https://t.co/01we34MeEr
🗓️ Poster: Wed, July 16 @ 4:30 pm PT | E‑2803

🧵

[sijia.liu @sijialiu17]

🚨 Excited to attend #ICML2025 and share our latest work (@OptML_MSU) on LLM unlearning -- think of it as AI surgery: removing harmful knowledge while preserving general utility. Catch us at:

🔹 [Paper 1] Tues, July 15 @ 4:30pm PT | E-1108
📄 Invariance Makes LLM Unlearning Resilient Even to Unanticipated Downstream Fine-Tuning [🔗 https://t.co/fb7sgRiTBc]
-- Even unrelated fine-tuning (e.g., math) can achieve invariance in unlearning via disentangled task vectors, which improves unlearning robustness against general post-unlearning fine-tuning operations.

🔹 [Paper 2] Wed, July 16 @ 4:30pm PT | E-2803
📄 Towards LLM Unlearning Resilient to Relearning Attacks: A Sharpness-Aware Minimization Perspective and Beyond [🔗 https://t.co/ERl2v8JsTQ]
-- Connecting adversarial unlearning to sharpness-aware optimization and general smoothness optimization

Also at #MUGen Workshop @ ICML:
🎤 Invited talk: Progress, Pitfalls & Prospects of LLM Unlearning
🧠 Oral: "Unlearning Isn’t Invisible: Detecting Unlearning Traces in LLMs from Model Outputs" [🔗 Long version: https://t.co/ZJwWX8Jc8v]
📌 Poster: "Reasoning Model Unlearning" [🔗 Long version: https://t.co/WJOcqb3O4r]

Grateful to my outstanding students (@ChongyuFan @wcsa23187 @yiwei_chen_ @Hi_Soumyadeep @zyh2022 @jia_jinghan) and wonderful collaborators from @MITIBMLab IBM (@NathalieBaraca1 Dennis Wei @p_ram_p) and Amazon (@Mingyi552237 @anil_k_ram) for their dedication and contributions to these works!

Looking forward to reconnecting with friends and meeting new ones--come say hi and chat about building safe, efficient, and trustworthy generative models! 🤝

[Pin-Yu Chen @pinyuchenTW]

(2/n) Oral Paper!
Task vector is one of the most popular methods for model editing and model merging. We provide a theoretical framework to understand its generalization and limitations.

With @LiHongkang_jntm @zyh2022 Shuai Zhang, Meng Wang @sijialiu17

https://t.co/tryRMx8Pg4

[sijia.liu @sijialiu17]

Check out our oral paper #ICLR25, led by @LiHongkang_jntm and @zyh2022, which shows when and why model editing via task vector stays probably effective, with application to LLM unlearning.

[Yihua Zhang @zyh2022]

@wcsa23187 Thanks Changsheng ☺️

[Yihua Zhang @zyh2022]

I'm thrilled to announce that I've been awarded the prestigious IBM PhD Fellowship 2024! @IBMResearch

A heartfelt thank you to my advisor, colleagues, and the @IBM award committee for their support and recognition.

#IBMPhDFellowship2024

https://t.co/a1BYOauGBm

[Yihua Zhang @zyh2022]

@LiangJeff95 I guess some reviews marked this paper with ethics concern, LoL. Never experienced this before, either.

[Yihua Zhang @zyh2022]

So grateful for such a babysitting-style tutorial! It took me ten minutes to read but I guess it will take me several months to digest. Thanks Manling!

[Manling Li @ManlingLi_]

If you are asked "Do you have any questions for me?" during the interview, here are some I like to ask:

Research:
- How do faculty members collaborate? Do you often apply for funding together?
- Is co-advising students a common practice here?
- Are there any plans or proposals for new research centers in the department/university?
- What does the dept care most in the next few years? Does the dept have a development plan for some specific research directions?
Mentoring:
- Do you have any suggestions to identify good students?
- Do students typically reach out for summer internships?
- What is the current TA-to-student ratio?
- Are there any new courses the department is considering offering?
- Do you have the faculty mentoring program?
Funding:
- How far in advance must proposals be submitted to the pre-award office (e.g., one week or three days)?
- Can technical parts of proposals be revised after submission to the pre-award office?
- What should I prepare for the administrative side of proposals?
- Are there any writing groups for career proposals?
- What funding sources do you or most faculties in the department often apply?
One all-purpose question:
- What do you like best about this university?

(My list centered around research, student mentoring, and funding projects, which I considered were my strengths and I would insert some talking points about my own related experience, but feel free to tailor your questions to your own talking points. I have heard that strong interviewees can skillfully showcase their expertise even through the questions they ask. I am far from that, but it is always good to focus on topics you feel most comfortable discussing.)

[Yihua Zhang @zyh2022]

🎉🎉🎉Who needs Santa when your boss is delivering wins like this? 🌟🌟🌟

[sijia.liu @sijialiu17]

Honored to receive the Amazon Research Award for advancing machine unlearning techniques--what a fantastic Christmas gift from @AmazonScience! Huge thanks to my hardworking students @OptML_MSU. Looking forward to new and fruitful collaborations with Amazon in 2025! @MSU_EGR

[sijia.liu @sijialiu17]

Honored to receive the Amazon Research Award for advancing machine unlearning techniques--what a fantastic Christmas gift from @AmazonScience! Huge thanks to my hardworking students @OptML_MSU. Looking forward to new and fruitful collaborations with Amazon in 2025! @MSU_EGR

[Yihua Zhang @zyh2022]

@VITAGroupUT @ccccrs_0908 @nvidia @TianlongChen4 What an amazing team!

[Yihua Zhang @zyh2022]

@ccccrs_0908 Well deserved! Congrats Ruisi!!!

[Yihua Zhang @zyh2022]

We do not really need to argue for a refund (and NeurIPS for sure will not refund either), this is just showing an attitude. Regarding the strategy, the email is the right one. See here: Cathy Francis is our Human Resource consultant. The hotline number and email are only for Human Resource questions, comments or violation of Code of Conduct. Do not call or email about registration or conference questions. The Human Resource email address is [email protected] and the phone number is 858-208-3810 for Human Resources.

[Yihua Zhang @zyh2022]

🚨 Call to Action: Demand Accountability and Request Refunds from NeurIPS! 🚨

🛑 To @NeurIPSConf: If you don't show your attitude, we will show ours! 🛑

We demand:
1️⃣ A full public apology from NeurIPS and the keynote speaker.
2️⃣ Refunds for all affected attendees.

❗️ Let’s make our voices heard! ❗

📧 Email your complaint (you can use the following template) to [email protected] and retweet this post to spread the word. Together, we can fight for accountability and inclusivity in academia.
#NeurIPS2024 #AccountabilityNow #AcademicIntegrity #StopRacismInAcademia

At @NeurIPSConf, the keynote speech featured discriminatory remarks targeting Chinese researchers. This act of blatant racial bias has deeply hurt the academic community. NeurIPS’ response, referring to it as mere "cultural generalization," falls far short of what is expected.

------> Complaint Email Template <-----

Subject: Formal Request for Refund and Public Apology Regarding Racist Remarks at NeurIPS 2024 Keynote Speech

Dear NeurIPS Organizing Committee,

I am writing to formally express my profound disappointment and outrage regarding the openly discriminatory and racially charged remarks made during the keynote speech at NeurIPS 2024. The incident in question is not merely a "cultural generalization," as described in your statement, but an explicit act of racial bias that has caused significant harm to the dignity and reputation of Chinese researchers and the broader academic community.

NeurIPS, as the organizer of this event, bears direct responsibility for this incident. The keynote speaker was invited by NeurIPS, making this incident a failure of due diligence and a breach of your own commitment to fostering an inclusive, safe, and respectful environment. This is a clear violation of NeurIPS' Code of Conduct and undermines the trust that the global research community has placed in your organization.

As a participant who paid a significant registration fee to attend NeurIPS under the expectation of a professional and inclusive environment, I demand the following actions:

1. A full refund of my registration fee, as the conference has failed to uphold its stated standards of inclusivity and professionalism.
2. A formal, public apology from both NeurIPS and the keynote speaker, acknowledging the harm caused by this incident and committing to actionable measures to prevent similar occurrences in the future.

If these demands are not met, we reserve the right to take further actions, including but not limited to:

1. Boycotting future NeurIPS events, urging other members of the academic community to do the same.
2. Filing a formal complaint to relevant authorities in Vancouver, where NeurIPS 2024 is hosted, to highlight the discriminatory nature of this incident and its potential legal and ethical violations.
3. Raising awareness across major academic and social media platforms to ensure accountability and transparency regarding this issue.

This incident has cast a long shadow over NeurIPS' reputation as a global leader in AI research. Failure to address this issue adequately will only deepen the harm done to the trust and inclusivity of the academic community. I urge NeurIPS to take immediate and decisive action to rectify this situation.

Sincerely,
[Your Full Name]
[Your Contact Information]

[Yihua Zhang @zyh2022]

“I thought adding some vague and evasive NOTE would make my racially discriminatory behavior seem less offensive.”

- Excuse given by MIT professor who is scheming and racist and now in the eye of the storm.

NOTE: Most MIT professors who I know are not discriminative or racist.

[Jiao Sun @sunjiao123sun_]

Mitigating racial bias from LLMs is a lot easier than removing it from humans!

Can’t believe this happened at the best AI conference @NeurIPSConf

We have ethical reviews for authors, but missed it for invited speakers? 😡

[Yihua Zhang @zyh2022]

@VITAGroupUT It is impressive this talented group remains so productive even if the boss is not around 😁

[sijia.liu @sijialiu17]

Excited to attend #NeurIPS2024 in Vancouver! 🎉

Check out the OPTML Menu of Innovations @ NeurIPS 2024:
🔗 https://t.co/4E3prMc84Q, featuring our latest research on adversarial ML and machine unlearning for generative AI.

Looking forward to reconnecting with old friends and meeting new ones. See you in Vancouver! 🌟 @OptML_MSU