The global ML scientific community will convene from July 6 to July 11, 2026 at the 43rd International Conference on Machine Learning (ICML) in Seoul, to present and discuss recent findings in the field. Principal Investigators, Postdoctoral Fellows as well as PhD Students from the groups of Michael Bronstein, Alex Tong and Ismail Ceylan at AITHYRA will present their latest accomplishments in formats such as Orals, Spotlights, Poster Presentations, and Workshops. Inter alia, Zander W. Blasingame (AITHYRA) will give a talk on a paper authored by him together with Chen Liu (Clarkson University) titled “Rex: A Family of (Stochastic) Runge-Kutta Solvers” on July 7 in the ICML Oral session on “2D AI for Science: Differential Equations & Molecular Modelling”.
In their paper Blasingame and Liu address the problem that modern AI image generators, protein designers, and molecular simulators work by gradually transforming random noise into a meaningful sample, one small step at a time. For many important uses - editing a real photograph, training the model from its outputs, or accurately measuring how likely a molecular shape is - it is necessary to run this process in reverse, recovering the original noise from a finished sample. In practice, however, tiny rounding errors pile up at every step, so the reverse trip never lands exactly where it started. Existing fixes are unstable, only mildly accurate, and almost none of them work when the generation process involves randomness. To solve this problem, Blasingame and Liu introduce Rex, a new family of numerical "step-takers" that are algebraically reversible: the backward step is the exact mathematical undo of the forward step, so the reconstruction is perfect by design - down to the precision of the computer's arithmetic. Rex works for both non-random and random generative models, and it is the reversible counterpart of many widely used solvers. Rex enables faithful image editing and more accurate sampling of molecular configurations - unlocking precision-critical applications that were previously out of reach.
In total, 6 papers with contributions from AITHYRA researchers were accepted at the ICML 2026 conference in Seoul.
Papers / Posters
(1) Rex: A Family of (Stochastic) Runge-Kutta Solvers
Zander W. Blasingame and Chen Liu
(2) Autoregressive Boltzmann Generators
Danyal Rehman, Charlie Tan, Yoshua Bengio, Joey Bose, Alexander Tong
(3) MacroGuide: Topological Guidance for Macrocycle Generation
Alicja Maksymiuk, Alexandre Duplessis, Ismail Ceylan, Alexander Tong, Fernanda Duarte, Michael Bronstein
(4) Riemannian Metric Matching for Scalable Geometric Modeling of Distributions
Jacob Bamberger, Adam Gosztolai, Pierre Vandergheynst, Michael Bronstein, Iolo Jones
(5) Categorical Flow Maps
Daan Roos, Oscar Davis, Floor Eijkelboom, Michael Bronstein, Max Welling, Ismail Ceylan, Luca Ambrogioni, Jan-Willem van de Meent
(6) Graph Neural Networks are not continuous across Graph Resolutions
Christian Koke, Abhishek Saroha, Yuesong Shen, Marvin Eisenberger, Bastian Rieck, Michael Bronstein, Daniel Cremers
#ICML2026