Clara Kümpel

Clara Kümpel

learning theory · neuroscience-inspired AI
Saarbrücken // Zürich // London

Hi, I'm Clara. I'm interested in how neural networks and humans learn. I like to apply ideas from neuroscience to machine learning theory and vice versa.

Recently, I am mostly interested in learning dynamics and optimisation to improve learning in neural networks from an efficiency, modular, neuroscience-inspired angle. I am a first-year ELLIS PhD student with Rebekka Burkholz at CISPA and Jared Tanner at the University of Oxford.

During my master's at ETH Zürich, I spent some time studying nonlinear learning at the Gatsby Unit (UCL) in Andrew Saxe's lab and self-supervised learning at Mila with Blake Richards' lab. I am also working on optimisation theory with Jonathan Cornford from the University of Leeds. Before I did a computer science degree with IBM, worked on neurosymbolic AI at the MIT-IBM Watson AI Lab, on computational neuroscience at IBM Research in Yorktown Heights (NY), and studied knowledge graphs and cloud computing in their research labs in Zurich and Stuttgart.

Feel free to reach out if you think we might get along :)

Interests optimisation learning theory comp. neuroscience self-supervised learning language & LLMs neuro-symbolic AI
News
  • September 2026Started my PhD! 🏄‍♀️
  • September 2026Attending the 13th Heidelberg Laureate Forum, bringing together young researchers with laureates in computer science and mathematics
  • July 2026oralI will be giving an Oral Presentation at ACL 2026 in San Diego
  • June 2026Finished my Master's at ETH Zürich in Computer Science
  • June 2026spotlightInvited Spotlight talk at the 7th International Conference on the Mathematics of Neuroscience and AI in Rome
  • March 2026posterPresenting a Poster at Cosyne 2026 in Lisbon
  • November 2025Visiting Andrew Saxe's lab at the Gatsby Unit in London at UCL!

Some recent work

·01It's Not What You Say, It's How You Say It: Evaluating LLM Responses to Expressions of BeliefACL 2026
·02Representation Geometry Governs Nonlinear LearningCOSYNE 2026
·03GLAudio Listens to the Sound of the GraphICML 2024 · workshop
It's Not What You Say, It's How You Say It: Evaluating LLM Responses to Expressions of Belief
ACL 2026 · main conference
Evaluating how large language models respond to how a belief is expressed — tone and framing — not just its propositional content.
Representation Geometry Governs Nonlinear Learning
COSYNE 2026
Comparing geometric and structural conditions for learning XOR in biological and artificial networks.
paper soon
GLAudio Listens to the Sound of the Graph
ICML 2024 · workshop
Geometry-grounded graph neural networks that propagate information like sound through a graph's structure.

Experience

2026 – now
ELLIS PhD Student
CISPA & University of Oxford · Rebekka Burkholz, Jared Tanner
2025 / 2026
Predoctoral Researcher
Gatsby Unit, UCL · Andrew Saxe
2025 – 2026
Research Assistant
Energy Policy Group, ETH Zürich
2024
Visiting Researcher
Mila, Montréal · Blake Richards
2022
Research Intern
IBM Thomas J. Watson Research Center
2021
Research Intern
MIT-IBM Watson AI Lab
2019 – 2022
IBM Corporate Student

Education

2023 – 2026
MSc Computer Science, Machine Intelligence
ETH Zürich
2022 – 2023
Preparatory Mathematics & CS studies
ETH Zürich
2019 – 2022
BSc Computer Science (corporate programme with IBM)
DHBW Stuttgart
Best Bachelor's Thesis Award, German Informatics Society

Awards & scholarships

2021 – 2026
Fellowship, German Academic Scholarship Foundation (Studienstiftung)
2023 – 2025
DAAD Scholarship for graduate studies at ETH Zürich
2025
Long-term visitor, Thematic Program on the Mathematics of Neuroscience
Fields Institute, Toronto
2025
Travel grant, School on Analytical Connectionism
UCL / Gatsby
2024
Invited young researcher, Heidelberg Laureate Forum (alumni 2025, 2026)
2023
Best Bachelor's Thesis Award
German Informatics Society
2023 – 2024
IDEA League Leadership Program, ETH representative

Community & art

Women in STEM

I strongly care about equal opportunities and am always happy to chat about issues, mentoring and advice :) I've mentored Girls in STEM and co-founded communities supporting women in STEM at ETH & in Zürich.

Art

I'm organizing a workshop on using art for science at the Heidelberg Laureate Forum. This year's exhibition ↗

Scientific community

I've organised workshops on art & science communication (@HLF) and the Free Energy Principle (for the Studienstiftung, the German Academic Scholarship Foundation), and have reviewed for CogSci 2026.