PeopleofSuperintelligence

Researcher

Chiyuan Zhang

Chiyuan Zhang is a machine-learning researcher whose work spans neural-network generalization, reinforcement learning and privacy. At MIT, he studied how semantic similarities between image categories can improve classification, then coauthored an ICLR Best Paper on random-label fitting and research on reinforcement-learning overfitting. He later examined how neural networks memorize training examples and coauthored LabelDP, a method that protects training labels when inputs are public; he also coauthored VaultGemma 1B, a differentially private Gemma-family model, in 2025. As of October 2026, he is a research scientist at Google DeepMind, builds self-improving agentic systems, and holds a PhD from CSAIL, MIT, earned while working with Tomaso Poggio at the Center for Brains, Minds and Machines.

Timeline

  1. 2015

    Learning with related labels

    In 2015, as an MIT graduate student in electrical engineering and computer science, he is a lead author on a paper about using semantic similarities between image categories to improve image classification.

  2. 2017

    An ICLR Best Paper Award

    In 2017, he coauthors Understanding deep learning requires rethinking generalization. Its experiments show that convolutional networks can fit random labels, and the paper wins an ICLR Best Paper Award.

  3. 2018

    Overfitting in deep reinforcement learning

    In 2018, he coauthors A Study on Overfitting in Deep Reinforcement Learning. The study finds that the same agents and algorithms can show very different test performance even when they achieve optimal training rewards.

  4. 2020

    Memorization and generalization

    In 2020, he coauthors What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimation. Its experiments show that memorization can benefit generalization on standard benchmarks.

  5. 2021

    Label differential privacy

    In 2021, he and Pasin Manurangsi present Deep Learning with Label Differential Privacy at NeurIPS. LabelDP treats inputs as public, protects training labels, and uses prior understanding of labels to improve model utility.

  6. By 2022

    Research scientist at Google Research

    By 2022, he is a research scientist at Google Research.

  7. 2025

    VaultGemma 1B

    In 2025, he coauthors VaultGemma 1B, a Gemma-family model trained with differential privacy.

  8. As of October 2026

    Research scientist at Google DeepMind

    As of October 2026, he is a research scientist at Google DeepMind and builds self-improving agentic systems. He holds a PhD from CSAIL, MIT, earned while working with Prof. Tomaso Poggio in the Center for Brains, Minds and Machines (CBMM).

Career

  1. NowGoogle DeepMindResearch scientist

Education

  1. MIT CSAILPhD

From the biography.

  1. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other universityIncludes Chiyuan Zhang. The Massachusetts Institute of Technology trained or employed 44 of the 187 AI major-prize winners, more than any other university. We compare doctorate and 5+ year employment links, career timing and current workplaces for 187 major-prize and 1,111 best-paper winners.
  2. Google had the most people shaping AI every year from 2014 to 2025Includes Google. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  3. Google has the highest combined h-index among AI organizationsIncludes Google. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.
  4. 20% of people at companies who shape AI also hold a university role, led by Google and Stanford UniversityIncludes Google. We estimate that 539 of the 6,766 people with a known current organization (8%) hold both a university role and a company role, including 20% of people at companies and 13% of people at universities.

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 Chiyuan Zhang yet. Send us a free-licensed photo.

  1. More-flexible machine learning news.mit.edu. Checked 22 October 2026.
  2. Understanding deep learning requires rethinking generalization arxiv.org. Checked 22 October 2026.
  3. UMN ML - 2021 Summer sites.google.com. Checked 22 October 2026.
  4. A Study on Overfitting in Deep Reinforcement Learning arxiv.org. Checked 22 October 2026.
  5. What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimation research.google. Checked 22 October 2026.
  6. Deep Learning with Label Differential Privacy research.google. Checked 22 October 2026.
  7. VaultGemma research.google. Checked 22 October 2026.
  8. Chiyuan Zhang pluskid.org. Checked 22 October 2026.