PeopleofSuperintelligence

Researcher

Hongyang Zhang

Also known as 张弘扬

Hongyang Zhang is a machine-learning researcher at the University of Waterloo whose work covers adversarial robustness, faster language-model inference and world models. After a master's degree at Peking University in 2015, he earned a PhD in Carnegie Mellon's Machine Learning Department in 2019, co-advised by Maria-Florina Balcan and David Woodruff. His TRADES defense, which grew out of a winning entry in the 2018 NeurIPS Adversarial Vision Challenge, appeared at ICML 2019, and in 2024 he coauthored EAGLE, a speculative sampling method for large language models. Google gave him a Research Scholar Award in 2025. He is a tenure-track assistant professor at Waterloo's Cheriton School of Computer Science, leads the SafeAI Lab, and is a faculty member at the Vector Institute.

Timeline

  1. 2015–2019

    Peking University to a CMU PhD

    He earns a master's degree at Peking University in 2015 and a PhD in the Machine Learning Department at Carnegie Mellon University in 2019. Maria-Florina Balcan and David P. Woodruff co-advise his doctorate.

  2. 2018–2019

    The TRADES defense method

    His team wins first place in the Robust Model and Targeted Attacks tracks of the NeurIPS 2018 Adversarial Vision Challenge in 2018. In 2019, he publishes TRADES at ICML with Yaodong Yu, Jiantao Jiao, Eric P. Xing, Laurent El Ghaoui and Michael I. Jordan, a defense that balances adversarial robustness against accuracy.

  3. 2021

    A postdoc at TTIC

    He is a postdoctoral fellow at the Toyota Technological Institute at Chicago in 2021, hosted by Avrim Blum and Greg Shakhnarovich. In April 2021, his team AdvRandom takes first place among 1,559 teams in the CVPR 2021 Security AI Challenger on unrestricted adversarial attacks on ImageNet.

  4. 2022–2024

    Teaching and awards at Waterloo

    In 2022, he teaches CS886 Robustness of Machine Learning and CS858 Security and Privacy of Machine Learning at Waterloo. He receives the Amazon Research Award and the WAIC Yunfan Award in 2023, and AAAI New Faculty Highlights in 2024.

  5. 2024

    EAGLE speculative sampling

    In 2024, he coauthors EAGLE with Yuhui Li, Fangyun Wei and Chao Zhang; the speculative sampling framework reaches a latency speedup of 2.7x to 3.5x on LLaMA2-Chat 70B. EAGLE-2, with dynamic draft trees, follows in June 2024 at 3.05x to 4.26x, and EAGLE is merged into vLLM, SGLang and NVIDIA's TensorRT-LLM.

  6. 2024–2025

    The Matrix and EAGLE-3

    In 2024, he coauthors The Matrix, a world simulator trained on AAA game data from Forza Horizon 5 and Cyberpunk 2077 that generates continuous 720p video streams with real-time control. EAGLE-3, released in March 2025, reaches speedups of up to 6.5x, about 1.4x over EAGLE-2.

  7. 2025

    Google award and Vector Institute

    He becomes a faculty member at the Vector Institute in 2025. In July 2025, Waterloo announces that he has received US$60,000 through the 2025 Google Research Scholar Program for a project on speeding up reasoning in large language models.

  8. As of October 2026

    SafeAI Lab at Waterloo

    He is a tenure-track Assistant Professor at Waterloo's Cheriton School of Computer Science, where he leads the SafeAI Lab, and a faculty member at the Vector Institute. He is invited to serve as Senior Area Chair for NeurIPS 2026 in March 2026, and EAGLE-4 is released in October 2026.

Career

  1. Since 2025Vector InstituteFaculty Member
  2. NowUniversity of WaterlooAssistant Professor
  3. 2021Toyota Technological Institute at ChicagoPostdoctoral Fellow

Education

  1. 2019Carnegie Mellon UniversityPhD
  2. 2015Peking UniversityMaster's degree

From the biography.

  1. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Vector Institute. Our estimates of where the 943 people shaping AI who worked at Google and Google DeepMind work now: 446 at Google and 452 across 278 other organizations.
  2. Nonprofit roles are 2.3 times as common in AI safetyIncludes Vector Institute. We estimate where people focused on safety, alignment and evaluation work. Nonprofit roles appear for 25.1% of safety people with known affiliations, compared with 11.0% in other areas.
  3. Universities (31%) and startups and other companies (25%) employ the largest shares of 2020–2026 entrantsIncludes Vector Institute. Among 819 people whose first recorded role started in 2020–2026 and who have a known current organization, 60% work in commercial sectors (25% at startups and other companies, 23% at frontier labs and 13% at big tech, with 9 in two), and 31% work at universities, compared with 66% for earlier entrants.
  4. The United States employs an estimated 48% of people shaping AIIncludes Vector Institute. We estimate where the people shaping AI work today, how universities, companies and research institutes split across countries, and how country shares have changed since the 1960s.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

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

  1. Hongyang Zhang | Cheriton School of Computer Science | University of Waterloo uwaterloo.ca. Checked 21 October 2026.
  2. Hongyang Zhang's Homepage hongyanz.github.io. Checked 21 October 2026.
  3. [1901.08573] Theoretically Principled Trade-off between Robustness and Accuracy arxiv.org. Checked 21 October 2026.
  4. Seminar • Machine Learning — New Advances in (Adversarially) Robust and Secure Machine Learning | Cheriton School of Computer Science | University of Waterloo uwaterloo.ca. Checked 21 October 2026.
  5. Hongyang Zhang - Vector Institute for Artificial Intelligence vectorinstitute.ai. Checked 21 October 2026.
  6. [2401.15077] EAGLE: Speculative Sampling Requires Rethinking Feature Uncertainty arxiv.org. Checked 21 October 2026.
  7. [2406.16858] EAGLE-2: Faster Inference of Language Models with Dynamic Draft Trees arxiv.org. Checked 21 October 2026.
  8. [2412.03568] The Matrix: Infinite-Horizon World Generation with Real-Time Moving Control arxiv.org. Checked 21 October 2026.
  9. [2503.01840] EAGLE-3: Scaling up Inference Acceleration of Large Language Models via Training-Time Test arxiv.org. Checked 21 October 2026.
  10. Chengnian Sun and Hongyang Zhang receive Google Research Scholar Program funding | Cheriton School of Computer Science | University of Waterloo cs.uwaterloo.ca. Checked 21 October 2026.