PeopleofSuperintelligence

Researcher

Mickel Liu

Also known as mickelliu

Mickel Liu is an AI researcher whose work spans multi-agent learning, safe reinforcement learning and language-model alignment. He coauthored MATE, BeaverTails, Safe RLHF and OmniSafe, then earned a computer science master's at Peking University before pursuing a PhD at the University of Washington. His 2025 coauthored paper Artificial Hivemind studies homogeneous language-model responses and won a NeurIPS Best Paper award in the Datasets & Benchmarks track; in 2026, he coauthored SPADE, a self-play framework for adaptive training environments. As of October 2026, he is a PhD student at UW's Paul G. Allen School and a visiting researcher at Meta FAIR.

Timeline

  1. 2022

    MATE for multi-agent learning

    In 2022, he coauthors MATE, a multi-agent environment for target-coverage control. The benchmark models camera and target agents in cooperative and competitive tasks.

  2. 2023

    The BeaverTails safety dataset

    In 2023, he coauthors BeaverTails, a dataset for safety alignment in large language models. It separates human annotations of helpfulness and harmlessness for question-answer pairs.

  3. 2024

    Safe RLHF and OmniSafe

    In 2024, he coauthors Safe RLHF, which separates human preferences for helpfulness and harmlessness, and OmniSafe, a framework for research on safe reinforcement learning.

  4. By 2025

    A master's and UW PhD

    By 2025, he is a second-year PhD student in the Paul G. Allen School of Computer Science & Engineering at the University of Washington and holds a computer science master's from Peking University.

  5. 2025

    Self-RedTeam at the Allen School

    At the Allen School's 2025 research showcase, he presents Self-RedTeam, an online self-play reinforcement-learning method for language-model safety. The method lets one model act as both attacker and defender.

  6. 2025

    Artificial Hivemind wins a NeurIPS award

    In 2025, he coauthors Artificial Hivemind, a study of repeated and homogeneous language-model responses built around the Infinity-Chat dataset of 26,000 open-ended queries. NeurIPS names it a Best Paper in the Datasets & Benchmarks track.

  7. 2026

    SPADE for adaptive training

    In 2026, he coauthors SPADE, a self-play reinforcement-learning framework where a single language model acts as an environment designer and reasoning agent. It generates executable, long-horizon training environments.

  8. As of October 2026

    PhD at UW, researcher at Meta

    As of October 2026, he is a PhD student at the University of Washington's Paul G. Allen School and a visiting researcher at Meta FAIR. He is co-advised by Natasha Jaques and Luke Zettlemoyer.

Career

  1. NowFAIR (Meta)Visiting Researcher

Education

  1. Peking UniversityMaster's in Computer Science
  2. University of WashingtonPhD

From the biography.

  1. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Meta. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.
  2. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes Meta. Our estimates of the organizations where 550 people worked directly before joining 9 frontier AI labs across 620 career transitions, comparing the main feeder organizations per lab.
  3. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Meta. 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.
  4. 22 of 133 people shaping AI with a former OpenAI role now work at AnthropicIncludes Meta. Our estimates of where OpenAI recruits came from and where 133 people with former OpenAI roles work now, including the companies and labs they founded.

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 Mickel Liu yet. Send us a free-licensed photo.

  1. MATE: Benchmarking Multi-Agent Reinforcement Learning in Distributed Target Coverage Control proceedings.neurips.cc. Checked 21 October 2026.
  2. BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Dataset proceedings.neurips.cc. Checked 21 October 2026.
  3. Safe RLHF: Safe Reinforcement Learning from Human Feedback proceedings.iclr.cc. Checked 21 October 2026.
  4. OmniSafe: An Infrastructure for Accelerating Safe Reinforcement Learning Research jmlr.org. Checked 21 October 2026.
  5. Annual Research Showcase: Innovations in Computing - Paul G. Allen School of Computer Science & Engineering cs.washington.edu. Checked 21 October 2026.
  6. Chasing Moving Targets with Online Self-Play Reinforcement Learning for Safer Language Models arxiv.org. Checked 21 October 2026.
  7. NeurIPS 2025 Awards neurips.cc. Checked 21 October 2026.
  8. Artificial Hivemind: The Open-Ended Homogeneity of Language Models (and Beyond) proceedings.neurips.cc. Checked 21 October 2026.
  9. SPADE: Self-Play in Adaptive Synthetic Executable Environments arxiv.org. Checked 21 October 2026.
  10. Mickel Liu mickel-liu.github.io. Checked 21 October 2026.
  11. mickelliu (Mickel Liu) · GitHub github.com. Checked 21 October 2026.