About

I'm a computer science PhD student at Stanford AI Lab. I'm grateful to be advised by Percy Liang and Sanmi Koyejo. I'm part of Stanford NLP, Stanford ML, and Stanford Trustworthy AI Research (STAIR). I also worked with Dan Boneh at Applied Cryptography Group and did AI research at Google DeepMind. My PhD is supported by the Stanford SoE Fellowship and the Amazon AI PhD Fellowship.

I work on language models, data & user privacy, and their intersection. Some past related projects: LLM data membership, machine unlearning, local-remote model collaboration, personalization, deploying distributed differentially private training to Android, and 1st-place at the US-UK Privacy-Enhancing Technologies Challenge.

Blog / Feedback / GitHub / Google Scholar / LinkedIn / Twitter

News and Olds

• Sept 2025: Teaching CS221 at Stanford with Percy; we are making a bunch of changes to reflect AI in modern times
• Aug 2025: Launched the UQ project, a new paradigm for AI evals. Help us verify AI solutions to unsolved problems! (paper)
• Aug 2025: Gave talks on pitfalls of LLM train set inclusion at Stanford, Berkeley, and OpenAI (slides)
• May 2025: Two papers accepted as Spotlights at ICML 2025: pitfalls of LLM train set inclusion and manipulating Chatbot Arena
• Oct 2024: Gave a guest lecture on intro to ML privacy at Northeastern University CS7375 (slides)
• Aug 2024: Gave three talks at Google DeepMind around LLM training set inclusion, privacy, and unlearning, respectively
• May 2024: Wrote a long post on machine unlearning (tweet); the field is rapidly evolving and clarity is much needed
  • Jun 2024: Press coverage and interview by Politico
  • May 2024: Top trending post on Hacker News
  • May 2024: Talked about unlearning as a guest on The Data Exchange Podcast
• May 2024: Best paper award at the ICLR'24 DPFM Workshop
• Aug 2023: Taught programming & algorithms in Ethiopia as part of AddisCoder 🇪🇹
• July 2023: Gave a talk at the NITRD Privacy R&D Interagency Working Group of the US government
• June 2023: Gave a talk about model personalization at SWIFT
• May 2023: Gave a talk about our entry to the PETs challenge at the Royal Society in London, UK
• Mar 2023: Led our awesome CMU team ("puffle") to win 1st place at the US-UK Privacy-Enhancing Technologies (PETs) Prize Challenge, Pandemic Forecasting Track (USD $120,000) (blog post at ML@CMU and extended version)
  • Mar 2023: See news by White House, UK Government, Summit for Democracy, DrivenData, NSF, and CMU
• Mar 2023: Our research on distributed DP (1, 2) is officially deployed to Android

Research

	UQ: Assessing Language Models on Unsolved Questions Fan Nie*, Ken Ziyu Liu*, Zihao Wang, Rui Sun, Wei Liu, Weijia Shi, Huaxiu Yao, Linjun Zhang, Andrew Ng, James Zou, Sanmi Koyejo, Yejin Choi, Percy Liang, Niklas Muennighoff* arXiv preprint UQ-Platform live at uq.stanford.edu PDF / BibTeX / Tweet @misc{nie2025uqassessinglanguagemodels, title = {UQ: Assessing Language Models on Unsolved Questions}, author = {Nie, Fan and Liu, Ken Ziyu and Wang, Zihao and Sun, Rui and Liu, Wei and Shi, Weijia and Yao, Huaxiu and Zhang, Linjun and Ng, Andrew Y. and Zou, James and Koyejo, Sanmi and Choi, Yejin and Liang, Percy and Muennighoff, Niklas}, year = {2025}, eprint = {2508.17580}, archivePrefix = {arXiv}, primaryClass = {cs.CL}, url = {https://arxiv.org/abs/2508.17580} }
	Language Models May Verbatim Complete Text They Were Not Explicitly Trained On Ken Ziyu Liu, Christopher A. Choquette-Choo*, Matthew Jagielski*, Peter Kairouz, Sanmi Koyejo, Percy Liang, Nicolas Papernot ICML 2025 Spotlight PDF / BibTeX / Slides / Poster / Tweet @article{liu2025language, title = {Language Models May Verbatim Complete Text They Were Not Explicitly Trained On}, author = {Liu, Ken Ziyu and Choquette-Choo, Christopher A and Jagielski, Matthew and Kairouz, Peter and Koyejo, Sanmi and Liang, Percy and Papernot, Nicolas}, journal = {arXiv preprint arXiv:2503.17514}, year = {2025} }
	Exploring and Mitigating Adversarial Manipulation of Voting-Based Leaderboards Yangsibo Huang, Milad Nasr, Anastasios Angelopoulos†, Nicholas Carlini†, Wei-Lin Chiang†, Christopher A Choquette-Choo†, Daphne Ippolito†, Matthew Jagielski†, Katherine Lee†, Ken Ziyu Liu†, Ion Stoica†, Florian Tramer†, Chiyuan Zhang† ICML 2025 Spotlight PDF / BibTeX @article{huang2025exploring, title={Exploring and mitigating adversarial manipulation of voting-based leaderboards}, author={Huang, Yangsibo and Nasr, Milad and Angelopoulos, Anastasios and Carlini, Nicholas and Chiang, Wei-Lin and Choquette-Choo, Christopher A and Ippolito, Daphne and Jagielski, Matthew and Lee, Katherine and Liu, Ken Ziyu and others}, journal={arXiv preprint arXiv:2501.07493}, year={2025} }
	Machine Unlearning Doesn't Do What You Think: Lessons for Generative AI Policy, Research, and Practice A Feder Cooper, Christopher A Choquette-Choo, Miranda Bogen, Matthew Jagielski, Katja Filippova, Ken Ziyu Liu, Alexandra Chouldechova, Jamie Hayes, Yangsibo Huang, Niloofar Mireshghallah, Ilia Shumailov, Eleni Triantafillou, Peter Kairouz, Nicole Mitchell, Percy Liang, Daniel E Ho, Yejin Choi, Sanmi Koyejo, Fernando Delgado, James Grimmelmann, Vitaly Shmatikov, Christopher De Sa, Solon Barocas, Amy Cyphert, Mark Lemley, Jennifer Wortman Vaughan, Miles Brundage, David Bau, Seth Neel, Abigail Z Jacobs, Andreas Terzis, Hanna Wallach, Nicolas Papernot, Katherine Lee arXiv 2024 PDF / BibTeX @article{cooper2024machine, title={Machine Unlearning Doesn't Do What You Think: Lessons for Generative AI Policy, Research, and Practice}, author={Cooper, A Feder and Choquette-Choo, Christopher A and Bogen, Miranda and Jagielski, Matthew and Filippova, Katja and Liu, Ken Ziyu and Chouldechova, Alexandra and Hayes, Jamie and Huang, Yangsibo and Mireshghallah, Niloofar and others}, journal={arXiv preprint arXiv:2412.06966}, year={2024} }
	Machine Unlearning in 2024 Ken Ziyu Liu An edcuational and position piece #1 trending post on Hacker News Blog Post / PDF version / BibTeX / Tweet / Hacker News / Podcast / Politico / Wall Street Journal @article{liu2024unlearning, title = {Machine unlearning in 2024}, author = {Liu, Ken Ziyu}, journal = {Stanford Computer Science}, year = {2024}, month = {May}, url = {https://ai.stanford.edu/~kzliu/blog/unlearning}, }
	Investigating Data Contamination for Pre-training Language Models Minhao Jiang, Ken Ziyu Liu, Ming Zhong, Rylan Schaeffer, Siru Ouyang, Jiawei Han, Sanmi Koyejo Tech Report Best Paper Award & Oral Presentation at DPFM @ ICLR'24 PDF / BibTeX / Code / Tweet @article{jiang2024investigating, title={Investigating Data Contamination for Pre-training Language Models}, author={Jiang, Minhao and Liu, Ken Ziyu and Zhong, Ming and Schaeffer, Rylan and Ouyang, Siru and Han, Jiawei and Koyejo, Sanmi}, journal={arXiv preprint arXiv:2401.06059}, year={2024} }
	Differentially Private Adaptive Optimization with Delayed Preconditioners Tian Li, Manzil Zaheer, Ziyu Liu, Sashank Reddi, Brendan McMahan, Virginia Smith ICLR 2023: International Conference on Learning Representations Oral Presentation at OPT 2022 @ NeurIPS'22 PDF / BibTeX / Code @misc{li2022differentially, title={Differentially Private Adaptive Optimization with Delayed Preconditioners}, author={Tian Li and Manzil Zaheer and Ken Ziyu Liu and Sashank J. Reddi and H. Brendan McMahan and Virginia Smith}, year={2022}, eprint={2212.00309}, archivePrefix={arXiv}, primaryClass={cs.LG} }
	On Privacy and Personalization in Cross-Silo Federated Learning Ziyu Liu, Shengyuan Hu, Zhiwei Steven Wu, Virginia Smith NeurIPS 2022: Conference on Neural Information Processing Systems (Camera-ready version available) Underpins 1st-place solution at UK-US PETs Challenge TPDP 2022 / PDF / Abstract / BibTeX / Code / Poster / Blog Post @inproceedings{liu2022privacy, title={On Privacy and Personalization in Cross-Silo Federated Learning}, author={Liu, Ziyu and Hu, Shengyuan and Wu, Zhiwei Steven and Smith, Virginia}, booktitle={Neural Information Processing Systems (NeurIPS)}, year={2022} }
	Motley: Benchmarking Heterogeneity and Personalization in Federated Learning Shanshan Wu, Tian Li, Zachary Charles, Yu Xiao, Ziyu Liu, Zheng Xu, Virginia Smith Preprint FL-NeurIPS'22 / PDF / BibTeX / Code @misc{wu2022motley, title={Motley: Benchmarking Heterogeneity and Personalization in Federated Learning}, author={Wu, Shanshan and Li, Tian and Charles, Zachary and Xiao, Yu and Liu, Ziyu and Xu, Zheng and Smith, Virginia}, year={2022}, eprint={2206.09262}, archivePrefix={arXiv}, }
	The Skellam Mechanism for Differentially Private Federated Learning Naman Agarwal†, Peter Kairouz†, Ziyu Liu† NeurIPS 2021: Conference on Neural Information Processing Systems Oral Presentation at PPML 2021 @ ACM CCS'21 Deployed to Android PDF / Abstract / BibTeX / Code / Talk 1 / Talk 2 / Poster / Slides @article{agarwal2021skellam, title={The Skellam Mechanism for Differentially Private Federated Learning}, author={Agarwal, Naman and Kairouz, Peter and Liu, Ziyu}, journal={Advances in Neural Information Processing Systems}, volume={34}, pages={5052--5064}, year={2021} }
	The Distributed Discrete Gaussian Mechanism for Federated Learning with Secure Aggregation Peter Kairouz†, Ziyu Liu†, Thomas Steinke† ICML 2021: International Conference on Machine Learning Oral Presentation at TPDP 2021 @ ICML'21 Deployed to Android Full PDF / Abstract / Short PDF / BibTeX / Code / Talk / Poster / Slides @inproceedings{kairouz2021distributed, title={The Distributed Discrete Gaussian Mechanism for Federated Learning with Secure Aggregation}, author={Kairouz, Peter and Liu, Ziyu and Steinke, Thomas}, booktitle={International Conference on Machine Learning}, pages={5201--5212}, year={2021}, organization={PMLR} }
	Learning Implicit Credit Assignment for Cooperative Multi-Agent Reinforcement Learning Meng Zhou*, Ziyu Liu*, Pengwei Sui, Yixuan Li, Yuk Ying Chung NeurIPS 2020: Conference on Neural Information Processing Systems RL Theory Workshop / PDF / BibTeX / Code @article{zhou2020learning, title={Learning implicit credit assignment for cooperative multi-agent reinforcement learning}, author={Zhou, Meng and Liu, Ziyu and Sui, Pengwei and Li, Yixuan and Chung, Yuk Ying}, journal={Advances in Neural Information Processing Systems}, volume={33}, pages={11853--11864}, year={2020} }
	Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition Ziyu Liu, Hongwen Zhang, Zhenghao Chen, Zhiyong Wang, Wanli Ouyang CVPR 2020: Conference on Computer Vision and Pattern Recognition Oral Presentation PDF / Abstract / Supp / BibTeX / Demo / Code / Star @inproceedings{liu2020disentangling, title={Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition}, author={Liu, Ziyu and Zhang, Hongwen and Chen, Zhenghao and Wang, Zhiyong and Ouyang, Wanli}, booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition}, pages={143--152}, year={2020} }

Teaching & Mentoring

I love teaching! Most recently, I was part of the teaching team of AddisCoder 2023 🇪🇹, an intensive summer school in Ethiopia for middle/high school students interested in programming and computer science. I helped with student admissions, created lab exercises, gave lab lectures, and graded exams. I was also the main IT guy responsible for managing 100+ lab machines and making sure students can do exercises under poor computing infrastructure.

I'm also involved in the following teaching/mentoring activities:

• Deep Learning Portal
• Student-Applicant Support Program @ Stanford
• CS Undergrad Mentoring Program @ Stanford
• Peer Mentoring Program @ USydney

While an undergrad at USyd, I was a teaching assistant (academic tutor) for the following classes:

• COMP9110 System Analysis and Modelling, 2020
• COMP3027 Algorithm Design, 2018
• SOFT2412 Agile Software Development Practices, 2019
• COMP2017 Systems Programming, 2018
• INFO1103 Intro to Programming, 2017

(On research mentoring)

Experience

Google DeepMind, Mountain View CA, United States Student Researcher, DeepMind Privacy/Security, better half of 2024
Stanford Artificial Intelligence Laboratory (SAIL), Stanford CA, United States PhD student, 2023-Present
Carnegie Mellon School of Computer Science, Pittsburgh PA, United States Research Assistant, (RI/MLD), 2021-2023
Google Research, (remote from Sydney) AI Resident Researcher, 2020-2021 (Left early for grad school deferred from 2020)
Facebook, Menlo Park CA, United States Software Engineer Intern, Messenger/Instagram Ranking, Winter 2019/2020 (Summer in 🦘🇦🇺)
Amazon Web Services, Sydney, Australia Software Engineer Intern, Safety Engineering, Winter 2018/2019 (Summer in 🦘🇦🇺)
Apple, Cupertino CA, United States Software Engineer Intern, Core OS, Summer 2018

Professional Service

• Reviewer for COLM 2025: Conference on Language Modeling
• Reviewer for ICML 2025: International Conference on Machine Learning
• Reviewer for TMLR: Transactions on Machine Learning Research
• Program Committee for Private ML-ICLR'24: Privacy Regulation and Protection in Machine Learning Workshop
• Reviewer for ICML 2024: International Conference on Machine Learning
• Reviewer for ICLR 2024: International Conference on Learning Representations
• Program Committee for FL-ICML'23: Workshop on Federated Learning and Analytics in Practice
• Reviewer for NeurIPS 2023: Conference on Neural Information Processing Systems
• Reviewer for ICCV 2023: International Conference on Computer Vision
• Reviewer for ICML 2023: International Conference on Machine Learning
• Reviewer for CVPR 2023: IEEE/CVF Conference on Computer Vision and Pattern Recognition
• Reviewer for AISTATS 2023: International Conference on Artificial Intelligence and Statistics
• Reviewer for NeurIPS 2022: Conference on Neural Information Processing Systems
• Reviewer for TIP 2022: IEEE Transactions on Image Processing
• Reviewer for ECCV 2022: European Conference on Computer Vision
• Reviewer for CVPR 2022: IEEE/CVF Conference on Computer Vision and Pattern Recognition
• Reviewer for IJCV 2021: International Journal of Computer Vision

Misc

• While not feeling the AGI, I try to read, travel, do olympic weightlifting, waltz (sometimes polka, lindy hop), bake cheesecakes, Dota 2, among other things
• I co-organized the weekly lunch/seminar for Stanford ML group
• I performed waltz/polka as part of the Stanford Viennese Ball 2025 Openning Committee
• Twitter, Mastodon, Bluesky, Instagram, Goodreads
• My Erdős number is 4 via three paths
• Consider using JAX! It's a beautiful thing
• Check out cs-sop.org if you're a prospective CS PhD applicant; I have benefited from this initiative and I have shared my statement there too

Visits since COVID: