Research
ELLIS Unit Freiburg is dedicated to advancing autonomous learning, uniting expertise in efficient learning, trustworthy ML, and physical AI.
Programs
Explore programs that our ELLIS unit offers — connecting researchers, providing training and mobility opportunities, and contributing to policy work.
Upcoming Events
We regularly organize official ELLIS/ELIZA social events in Freiburg. This involves short talks from ELLIS/ELIZA students followed by discussion starting at 17:30 usually, following a BBQ at the Technische Fakultät campus or a visit to a pub, providing the perfect opportunity for networking. The location for the talks is the seminar room of Building 080. If you want to participate in the next event, please send an email to the organizers to be added to the official mailing list.
Past Events
April 1st, 2026
•
IMBIT, NEXUS LAB, Georges-Köhler-Allee 201, 79110 Freiburg
Revisiting the Platonic Representation Hypothesis: An Aristotelian View
The Platonic Representation Hypothesis suggests that representations from neural networks are converging to a common statistical model of reality. We show that the existing metrics used to measure representational similarity are confounded by network scale: increasing model depth or width can...
Guest Lecture
March 24th, 2026
•
Robot Learning Lab; Seminar Room, Georges-Köhler-Allee 080, 79110 Freiburg
From 2D Similarities to 3D Correspondences
Self-supervised visual representations often capture semantic part similarities, features that respond consistently to wings, legs, or wheels. However, these features encode 2D similarities rather than 3D correspondences. To reason in a 3D world, an ideal representation would provide 3D...
Guest Lecture
February 19th, 2026
•
Robot Learning Lab; Seminar Room, Georges-Köhler-Allee 080, 79110 Freiburg
Accelerating 3D Solar atmosphere modelling with Graph Neural Networks
Simulating the sun's atmosphere requires solving complex non-local, non-linear systems of coupled equations. A massive computational bottleneck that typically restricts researchers to 1.5D approximations or requires millions of CPU hours for full 3D snapshots. We present a solution that bypasses...
Guest Lecture
