Siyuan Guo
AI Scientist, Prior Labs
I am a machine learning researcher trying to understand intelligence. I am currently an AI scientist at Prior Labs, where I work on scaling laws and foundation models for causal inference. In 2026 I was named to the Forbes 30 Under 30 Europe list for Science & Healthcare. I also serve as an Area Chair for ICLR 2027. I did my PhD at the University of Cambridge with Ferenc Huszár and at the Max Planck Institute for Intelligent Systems with Bernhard Schölkopf, and interned at Meta FAIR in New York.
My long-term interest is AI + Science: building a physics-inspired mathematical theory of intelligence, and using AI to accelerate scientific discovery.
Research
- How can principles from nature guide the design of learning systems? This underpins my physics of learning programme: learning, too, follows a least-action principle, and classical learning algorithms are derivable from it.
- How can we identify fundamental laws of nature directly from data? This includes the causal de Finetti line of work on the Bayesian foundations of causality, and foundation models for in-context causal inference in scientific applications.
Collaborations. I am always happy to connect with people working on science of AI and / or AI for science. Contact me at sguo26v@gmail.com.
News
- Jul 2026Invited Speaker at the ICML AI4Math workshop in Seoul on the physics of learning.
- May 2026Prior Labs was acquired — an exciting time at the intersection of academia and industry.
- Apr 2026Named to the Forbes 30 Under 30 Europe list for Science & Healthcare. [LinkedIn]
- Feb 2026Invited talk at CISPA Helmholtz Center: Physics of Learning and Structure.
- Feb 2026Invited talk at MBZUAI: Physics of Learning and Structure.
- Jan 2026Invited talk at Lawrence Berkeley National Laboratory: Physics of Learning and Structure.
Earlier news
- Dec 2025Research visit at the Flatiron Institute, NYC.
- Nov 2025Invited talk at kausable.ai: Physics of Learning and Do-PFN.
- Oct 2025Invited Friday Talk at Tübingen: Physics of Learning.
- Oct 2025Fireside chat with Ploutos.ai on the physics of learning. [Stream]
- Oct 2025Invited talk at the Gatsby Computational Neuroscience Unit, UCL.
- Sep 2025Preprint: Physics of Learning on arXiv.
- Sep 2025Selected as an MIT EECS Rising Star.
- Sep 2025NeurIPS spotlight: Do-PFN (top 3.19%).
- Jul 2025Opening lecturer at the Cambridge ELLIS Summer School for Causality Day.
- May 2025Received the G-Research PhD Prize.
- Jan 2025ICLR spotlight: Identifiable Exchangeable Mechanisms (top 5.1%).
- Sep 2024NeurIPS oral: Do Finetti (top 0.46%).
- Jul 2024Joined Meta FAIR, NYC as a research scientist intern.
- Jul 2024Invited lecturer at the Cambridge ELLIS Summer School.
- Jul 2024Won the MPI-IS Outstanding Female Doctoral Student Prize.
Students & alumni
- Yuche Gao Co-advised with J. M. Hernández-Lobato PerturbPFN for Drug Perturbation, FMSD@NeurIPS 2026
- Anna Kerekes Co-mentored with B. Schölkopf
- Michelle Chao Chen Co-advised with B. Schölkopf Structural Information in LLMs, ACL 2026
- Bálint Mucsányi Co-mentored with P. Reizinger Identifiable Representations for RL, ICLR 2026
- Moritz Miller Co-advised with J. Peters and B. Schölkopf In-context Counterfactual Reasoning, NeurIPS 2025
- Szilvia Ujváry Co-mentored with F. Huszár
Selected publications
- Physics of Learning: A Lagrangian Perspective to Different Learning Paradigms arXiv:2509.21049, 2025 Cited in Forbes 30 Under 30
- Do-PFN: In-Context Learning for Causal Effect Estimation NeurIPS, 2025 Spotlight · top 3.2%
- Do Finetti: On Causal Effects for Exchangeable Data NeurIPS, 2024 Oral · top 0.5%
- Causal de Finetti: On the Identification of Invariant Causal Structure in Exchangeable Data NeurIPS, 2023 · code
* equal contribution · † equal supervision · full list on Google Scholar