Shengran Hu

Hello! I am Shengran Hu (胡圣然). I am a Computer Science Ph.D. student at University of British Columbia, advised by Prof. Jeff Clune. Previously, I was a research scientist intern at Sakana AI.

I’m interested in how patterns and knowledge emerge from complex systems like natural evolution and scientific discovery. I study this by building open-ended, agentic AI systems that can accumulate complexity in language.

Email / Google Scholar / Github / Twitter

Selected Publications

(*denotes equal contribution †denotes equal advisory)

Learning to Continually Learn via Meta-learning Agentic Memory Designs

Yiming Xiong, Shengran Hu, and Jeff Clune†

arXiv preprint arXiv:2602.07755 , 2026

@article{zhang2025darwin,
  title = {Learning to Continually Learn via Meta-learning Agentic Memory Designs},
  author = {Xiong, Yiming and Hu, Shengran and Clune†, Jeff},
  journal = {arXiv preprint arXiv:2602.07755},
  year = {2026},
  x = {https://x.com/jeffclune/status/2021242681826095179}
}

Darwin Gödel Machine: Open-Ended Evolution of Self-Improving Agents

Jenny Zhang*, Shengran Hu*, Cong Lu, Robert Lange†, and Jeff Clune†

International Conference on Learning Representations (ICLR), 2026

@article{zhang2025darwio,
  title = {Darwin Gödel Machine: Open-Ended Evolution of Self-Improving Agents},
  author = {Zhang*, Jenny and Hu*, Shengran and Lu, Cong and Lange†, Robert and Clune†, Jeff},
  journal = {International Conference on Learning Representations},
  year = {2026},
  x = {https://x.com/SakanaAILabs/status/1928272612431646943}
}

Automated Design of Agentic Systems

Shengran Hu, Cong Lu, and Jeff Clune

International Conference on Learning Representations (ICLR), 2025

🏆 Outstanding Paper (NeurIPS 2024 Open-World Agent Workshop)

@article{hu2024ADAS,
  title = {Automated Design of Agentic Systems},
  author = {Hu, Shengran and Lu, Cong and Clune, Jeff},
  journal = {International Conference on Learning Representations},
  year = {2025},
  x = {https://x.com/jeffclune/status/1825551351746867502}
}

The AI Scientist-v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Search

Yutaro Yamada*, Robert Lange*, Cong Lu*, Shengran Hu, Chris Lu, Jakob Foerster, Jeff Clune†, and David Ha†

arXiv preprint arXiv:2504.08066 , 2025

@article{Yamada2025AISCIv2,
  title = {The {AI} Scientist-v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Search},
  author = {Yamada*, Yutaro and Lange*, Robert and Lu*, Cong and Hu, Shengran and Lu, Chris and Foerster, Jakob and Clune†, Jeff and Ha†, David},
  journal = {arXiv preprint arXiv:2504.08066},
  year = {2025},
  x = {https://x.com/SakanaAILabs/status/1911603439832080635}
}

Automated Capability Discovery via Foundation Model Self-Exploration

Cong Lu*, Shengran Hu*, and Jeff Clune

arXiv preprint arXiv:2502.07577 , 2025

@article{lu2025ACD,
  title = {Automated Capability Discovery via Foundation Model Self-Exploration},
  author = {Lu*, Cong and Hu*, Shengran and Clune, Jeff},
  journal = {arXiv preprint arXiv:2502.07577},
  year = {2025},
  x = {https://x.com/jeffclune/status/1889568685632667672}
}

Thought Cloning: Learning to Think while Acting by Imitating Human Thinking

Shengran Hu, and Jeff Clune

Advances in Neural Information Processing Systems (NeurIPS), 2023

Spotlight (top 3.1% in 12,343)

Language is often considered a key aspect of human thinking, providing us with exceptional abilities to generalize, explore, plan, replan, and adapt to new situations. However, Reinforcement Learning (RL) agents are far from human-level performance in any of these abilities. We hypothesize one reason for such cognitive deficiencies is that they lack the benefits of thinking in language and that we can improve AI agents by training them to think like humans do. We introduce a novel Imitation Learning framework, Thought Cloning, where the idea is to not just clone the behaviors of human demonstrators, but also the thoughts humans have as they perform these behaviors. While we expect Thought Cloning to truly shine at scale on internet-sized datasets of humans thinking out loud while acting (e.g. online videos with transcripts), here we conduct experiments in a domain where the thinking and action data are synthetically generated. Results reveal that Thought Cloning learns much faster than Behavioral Cloning and its performance advantage grows the further out of distribution test tasks are, highlighting its ability to better handle novel situations. Thought Cloning also provides important benefits for AI Safety and Interpretability, and makes it easier to debug and improve AI. Because we can observe the agent’s thoughts, we can (1) more easily diagnose why things are going wrong, making it easier to fix the problem, (2) steer the agent by correcting its thinking, or (3) prevent it from doing unsafe things it plans to do. Overall, by training agents how to think as well as behave, Thought Cloning creates safer, more powerful agents.
@article{hu2023ThoughtCloning, title = {{Thought Cloning}: Learning to Think while Acting by Imitating Human Thinking}, author = {Hu, Shengran and Clune, Jeff}, journal = {Advances in Neural Information Processing Systems}, volume = {36}, year = {2023}, x = {https://x.com/jeffclune/status/1664618665160085505} }

Shengran Hu

Selected Publications

(*denotes equal contribution †denotes equal advisory)

Selected Media Coverage

Selected Talks