ETH-EPFL Summer School
Reliable ML 2026
Reliable Machine Learning: from LLMs to cyber-physical and biological Systems
Scroll to explore
Reliable Machine Learning: from LLMs to cyber-physical and biological Systems
Interacting systems—ranging from biological networks and social webs to critical infrastructures like power grids—pose distinctive modeling challenges. Unlike many physical systems with well-established governing equations, most interacting systems lack explicit dynamical laws, making data-driven modeling and machine learning essential.
Yet, standard ML methods often break down under distribution shifts or high noise, and struggle to provide reliable predictions in the face of unexpected or rare scenarios, especially in the context of complex dynamical systems.
How can we design AI methods that are robust, reliable, and generalizable when learning from and acting on evolving, intelligent systems under distributional shifts?
Reliable ML 2026 brings together researchers in machine learning, distributional robustness, causal learning, network modeling, robotics, and feedback control systems to address this core question.
Cyber-physical reliability of electric power networks
Robust gene regulation network inference for cancer drug response
Trustworthy multi-agent large language models
Reliable robot control in changing environments
World-class experts in robust AI, causality, and intelligent systems
EPFL
Distributional Robustness in ML
ETH Zürich
Safe and Efficient Exploration in Model-Based Reinforcement Learning
ETH Zürich
Learning under Change
Apple
Causality and Distribution Shift
University of Tübingen
Continual Learning under Distribution Shift
Apple Research
Uncertainty and Exploration in LLMs and Agents
Isomorphic Labs
Representation Learning for Robustness
Nomagic
Specialist and Generalist Physical AI
Cornell University
Learning under Local and Global Corruptions
Northwestern University
On the Robustness Trade-off: Designing Optimal DRO Formulations
EPFL & Xent Labs
AGI Meta-Game Objective: A Derivation
EPFL & Apple Research
Dynamic Reasoning and Planning
Lectures, hands-on sessions, panels, and networking opportunities
Establishing the theoretical groundwork for reliable AI systems
Organizers
Andreas Krause (ETH Zürich)
Florian Tramèr (ETH Zürich & Invariant Labs)
Daniel Kuhn (EPFL)
Markus Wulfmeier (Nomagic)
Panel Discussion
Understanding and adapting to changing data distributions
Claire Vernade (University of Tübingen)
Emmanuel Abbé (EPFL)
Jonas Peters (ETH Zürich)
Organizers
From theory to implementation in complex systems
Christina Heinze-Deml (Apple)
Jörn Jacobsen (Isomorphic Labs)
Soroosh Shafiee (Cornell University)
Amine Bennouna (Northwestern University)
Organizers
Looking ahead: future challenges and synthesis
Clément Hongler (EPFL & Xent Labs)
Michael Kirchhof (Apple)
Organizers
Located in the heart of Zürich, Switzerland
Tannenstrasse 3, Room F34
8006 Zürich, Switzerland
The summer school will be held at ETH Zürich in the city center, overlooking the beautiful old town and Lake Zürich.
Zürich Airport (ZRH) is Switzerland's largest international airport with direct connections to most major cities worldwide. Located approximately 10 km north of the city center.
Zürich Hauptbahnhof (HB) is one of Europe's best-connected railway hubs.
Book via sbb.ch for "Supersaver" early-bird discounts.
The iconic funicular takes you directly from Central square to ETH's main building in under 2 minutes. Free with any ZVV ticket.
5 min walk from HB + 2 min ride
Take Tram 6 (direction Zoo) or Tram 10 (direction Zürichberg) from Zürich HB to "ETH/Universitätsspital".
~5 min from Zürich HB
A scenic 15-minute uphill walk from Zürich HB through the university quarter.
~15 min walk (uphill)
Join us for an evening of good food and great conversation
Sihlstrasse 28, 8001 Zürich
Founded in 1898, Haus Hiltl located right in the heart of Zürich, offers a buffet-style cuisine (over 100 homemade dishes daily) that blends Swiss, Mediterranean, and Indian influences. We'll gather here for the summer school's social dinner.
Switzerland is part of the Schengen Area. Depending on your nationality, you may need a Schengen visa to attend.
If you require an invitation letter for your visa application, please contact us at ramzi.dakhmouche@epfl.ch after registering.
We recommend starting the visa process at least 3 months in advance.
Shared accommodation will be arranged for participants who apply for financial aid.
Special hotel rates would be provided for the rest.
More details will be provided upon registration confirmation.
Zürich in August is typically warm and pleasant, with average temperatures of 18–26°C (64–79°F). Occasional rain showers are possible.
Eduroam WiFi is available throughout ETH Zürich campus. If your institution participates in eduroam, no additional setup is needed.
Guest WiFi credentials will be provided at registration for those without eduroam access.
Included: Coffee breaks (apéros) and social event dinner.
Not included: Lunch is on your own. There are many affordable options on and near campus.
Please indicate any dietary requirements during registration.
Registration will open in Spring 2026. Leave your details to be notified when registration opens.
Included: Apéros, social event dinner, and special hotel rates.
Not included: Lunch (on your own).
Course credits: Students will earn 2 ECTS equivalent credits.
Master students, PhD students, and young postdocs in:
Presenting a poster will increase your chances of acceptance.
EPFL-Empa · PhD Student
Uncertainty quantification and robustness for LLMs and network systems.
EPFL · PhD Student
Large-scale optimization and decision-making under uncertainty.
ETH Zürich · PhD Student
Safe learning, multi-agent systems, and sequential decision-making.
ETH Zürich · Postdoc
Learning and adaptive systems, machine learning.
For questions about Reliable ML 2026, please reach out to us.
