PeopleofSuperintelligence

Researcher

Yunhao Tang

Also known as Robin Tang

Yunhao Tang is a reinforcement learning researcher whose career runs through DeepMind, Llama research, Mistral and Anthropic. They began in physics, worked on reinforcement learning through a PhD at Columbia University, and then spent a few years at DeepMind in London and Paris with Rémi Munos. In 2023, they became a core contributor to Gemini and, in 2024, co-authored Generalized Preference Optimization. For Llama research, they spearheaded online RL and helped scale the training to Llama 3.3 in 2024 and Llama 4 in 2025; at Mistral, they co-authored the Magistral reasoning model in 2025. They are a member of technical staff on the pre-training team at Anthropic.

Now at
Anthropic
Badges

Timeline

  1. By 2019

    Columbia PhD in reinforcement learning

    After first studying physics, they work on reinforcement learning through a PhD at Columbia University. In 2019, they co-author "Reinforcement Learning for Integer Programming: Learning to Cut" with Shipra Agrawal and Yuri Faenza, which trains a deep RL agent to select cutting planes in integer programming solvers.

  2. 2020

    Search as policy optimization

    In 2020, they co-author "Monte-Carlo Tree Search as Regularized Policy Optimization" with Jean-Bastien Grill, Rémi Munos and others. The paper shows that AlphaZero's search heuristics approximate the solution of a regularized policy optimization problem, and it proposes an AlphaZero variant that uses the exact solution.

  3. By 2023

    Research scientist at DeepMind

    At DeepMind in London and Paris, they spend a few years researching deep RL algorithms and systems with Rémi Munos. In April 2023, they present a seminar on self-predictive representation learning for reinforcement learning as a research scientist at Google DeepMind in London.

  4. 2023–2024

    Core contributor to Gemini

    In December 2023, Google DeepMind launches Gemini, and they are a core contributor, also listed among the authors of the technical report that introduces the Ultra, Pro and Nano sizes. Their work covers Gemini v1 through 1.5 post-training, with a focus on tool use and agents.

  5. December 2023–February 2024

    Generalized preference optimization

    In February 2024, Generalized Preference Optimization (GPO) appears with them as first author; its family of offline losses includes DPO, IPO and SLiC as special cases. They are also an author of Nash Learning from Human Feedback, posted in December 2023, which fine-tunes language models toward the Nash equilibrium of a preference model.

  6. 2024–2025

    An RL stack for Llama

    At Llama research, they spearhead the prototype and algorithmic recipes for online RL and, as part of a small team, scale the training to Llama 3.3 and Llama 4. Llama 3.3 arrives in December 2024 as the first Llama model trained with a large-scale RL stack, and Llama 4 follows in April 2025. The 2025 paper LlamaRL, which they co-author, presents a fully distributed, asynchronous RL framework used in Llama 3 post-training.

  7. June 2025

    Co-authoring Magistral

    At Mistral, they research reasoning, and in June 2025 they are among the authors of Magistral, Mistral's first reasoning model. The Magistral paper describes a scalable RL pipeline built on the company's own models and infrastructure. Magistral Medium is trained for reasoning with RL alone, and Magistral Small is open-sourced under Apache 2.0.

  8. As of October 2026

    Pre-training at Anthropic

    As of October 2026, they are a member of technical staff on the pre-training team at Anthropic, with interests in training and reinforcement learning.

Career

  1. NowAnthropicMember of Technical Staff
  2. 2025Mistral AI
  3. Google DeepMindResearch Scientist

Education

  1. 2019Columbia UniversityPhD

From the biography.

  1. The median h-index is 48 at universities and 14 at frontier AI labsIncludes Yunhao Tang. The median h-index is 48 at universities and 14 at frontier AI labs. We compare publishing careers, recent citations, dual appointments and job moves across 2,420 university researchers and 308 frontier-lab researchers.
  2. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Yunhao Tang. 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.
  3. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Yunhao Tang. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.
  4. 16% of AI career moves in the 2020s went from a university to a companyIncludes Yunhao Tang. We estimate that 16% of AI career moves in the 2020s went from a university to a company; 29% of moves starting at a university went to a company, and 76% ended the university role.

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 Yunhao Tang yet. Send us a free-licensed photo.

  1. Yunhao (Robin) Tang robintyh1.github.io. Checked 21 October 2026.
  2. Reinforcement Learning for Integer Programming: Learning to Cut arxiv.org. Checked 21 October 2026.
  3. Monte-Carlo Tree Search as Regularized Policy Optimization arxiv.org. Checked 21 October 2026.
  4. AIDRC Seminar Series - Yunhao Tang tii.ae. Checked 21 October 2026.
  5. Gemini: A Family of Highly Capable Multimodal Models arxiv.org. Checked 21 October 2026.
  6. News - Yunhao (Robin) Tang robintyh1.github.io. Checked 21 October 2026.
  7. Generalized Preference Optimization: A Unified Approach to Offline Alignment arxiv.org. Checked 21 October 2026.
  8. Nash Learning from Human Feedback arxiv.org. Checked 21 October 2026.
  9. LlamaRL: A Distributed Asynchronous Reinforcement Learning Framework for Efficient Large-scale LLM Training arxiv.org. Checked 21 October 2026.
  10. Models - Yunhao (Robin) Tang robintyh1.github.io. Checked 21 October 2026.
  11. Magistral arxiv.org. Checked 21 October 2026.