Conference Agenda

Lost (and found) in latent land: applications to weather and climate

Pierre Gentine

Columbia University, United States of America

Embracing the Tyranny of Testing

Moritz Hardt

Max Planck Institute for Intelligent Systems, Germany

Bridging Perception and Logic: An Abductive Learning Cycle for Semantically Anchored Facial Expression Recognition

Jens Johannes Gebele^1,2, Philipp Brune¹, Frank Schwab², Sebastian von Mammen²

1: Neu-Ulm University of Applied Sciences, Germany; 2: University of Würzburg

Human-in-the-loop Concept Discovery and Curation in Vision Foundation Models

Hyesu Lim¹, Jinho Choi², Jaegul Choo², Steffen Schneider¹

1: Helmholtz Munich, Germany; 2: KAIST AI, South Korea

TACTIC: Tabular-Attribute Conditioned Transformer for Image Classification

Laura Daza^1,2, Marta Hasny^1,2, Cristina González^1,2, Julia Schnabel^1,2,3

1: Helmholtz Munich, Germany; 2: Technical University of Munich, Germany; 3: King's College London, UK

Learning Physical Geometry from Noisy Helical Particle Tracks: A Comparative Study of Transformers, SBI, and JEPA

Sylvain Caillou, Jan Stark, Alexis Vallier, Catherine Biscarat

L2I Toulouse, CNRS/IN2P3, Université de Toulouse

MLflow pilot service for Helmholtz researchers

Lisana Berberi, Christophe Laures, Khadijeh Alibabaei, Valentin Kozlov, Achim Streit

Karlsruhe Institute of Technology (KIT), Germany

Towards Robust Foundation Models for Digital Pathology

Jonah Kömen^1,2, Edwin de Jong³, Julius Hense^1,2, Hannah Marienwald^1,2, Jonas Dippel^1,2,3, Philip Naumann^1,2, Eric Marcus⁴, Lukas Ruff³, Maximilian Alber^3,7, Jonas Teuwen⁴, Frederick Klauschen^1,5,6,7, Klaus-Robert Müller^1,2,8,9

1: Berlin Institute for the Foundations of Learning and Data (BIFOLD), Berlin, Germany; 2: Machine Learning Group, Technische Universität Berlin, Berlin, Germany; 3: Aignostics, Berlin, Germany; 4: The Netherlands Cancer Institute Amsterdam (NKI), Antoni van Leeuwenhoek Hospital (AvL), Amsterdam, Netherlands; 5: Institute of Pathology, Ludwig Maximilian University, Munich, Germany; 6: German Cancer Research Center, Heidelberg, and German Cancer Consortium, Munich, Germany; 7: Institute of Pathology, Charité Universitätsmedizin, Berlin, Germany; 8: Department of Artificial Intelligence, Korea University, Seoul, Korea; 9: Max-Planck Institute for Informatics, Saarbrücken, Germany

Sailing Past Syntax: A Human-in-the-Loop Framework for Safe Generative AI in Science

Nico Schuster

Centre de Physique des Particules de Marseille, France

The Helmholtz Model Zoo: Enabling AI Model Sharing and Inference in the Helmholtz Cloud

Hans Werners, Engin Eren, Philipp Heuser

Deutsches Elektronen-Synchroton DESY, Germany

Improving Reliability of LLM-Based Robotic Task Planning Through Domain Adaptation and Benchmarking

Kanak Pandit, Ananya Kulkarni, Muhammad Saeed

ARENA2036, Germany

Learning Actionable Insights from Scientific Data

Jilles Vreeken

CISPA Helmholtz Center for Information Security, Germany

RIBEX: Predicting and Explaining RNA Binding Across Structured and Intrinsically Disordered Regions (IDR)-rich Proteins

Samuele Firmani^1,2, Felix Steinbauer^2,4, Gjergji Kasneci^2,3,4, Annalisa Marsico¹, Marc Horlacher^1,2

1: Helmholtz Munich, Germany; 2: School of Computation, Information and Technology, TUM; 3: School of Social Sciences and Technology, TUM; 4: Munich Center for Machine Learning (MCML)

FoldSAE: Learning to Steer Protein Folding Through Sparse Representations

Wojciech Zarzecki^1,2,4, Paulina Szymczak³, Ewa Szczurek^1,3, Kamil Deja^2,4

1: University of Warsaw, Poland; 2: Warsaw University of Technology, Poland; 3: Helmholtz Munich, Germany; 4: Ideas Research Institute

Interpreting Multimodal Latent Spaces in Single-Cell Multi-omics with Feature Attribution Techniques

Anis Ismail, Alejandro Sifrim

Laboratory of Multi-omics Integrative Bioinformatics, Department of Human Genetics, KU Leuven, Leuven, Belgium.

Embpy and Perturbation World Models: A Foundation-Model Stack for Predicting Cellular Responses

Gonçalo Pinto

Helmholtz Munich, Germany

Uncertainty-Guided Generation of Dark-Field Radiographs

Lina Felsner^1,2,3, Henriette Bast^4,5,6, Tina Dorosti^4,5,6, Florian Schaff^4,5,6, Franz Pfeiffer^4,5,6,7, Daniela Pfeiffer^5,6,7, Julia Schnabel^1,2,3,8

1: School of Computation, Information and Technology, Technical University of Munich, 85748 Garching, Germany; 2: Munich Center for Machine Learning (MCML), Munich, Germany; 3: Institute of Machine Learning in Biomedical Imaging, Helmholtz Munich, 85764 Neuherberg, Germany; 4: Department of Physics, School of Natural Sciences, Technical University of Munich, 85748 Garching, Germany; 5: Munich Institute of Biomedical Engineering, Technical University of Munich, 85748 Garching, Germany; 6: Institute for Diagnostic and Interventional Radiology, School of Medicine and Health, TUM University Hospital Klinikum rechts der Isar, Technical University of Munich, 81675 Munich, Germany; 7: Institute for Advanced Study, Technical University of Munich, 85748 Garching, Germany; 8: School of Biomedical Engineering & Imaging Sciences, King’s College, London, UK

AI-assisted Classification of Polar Phytoplankton’s Functional Biodiversity using Multispectral Imaging Flow Cytometry

Natasha Bryan¹, Anna Naberhaus¹, Sarah Lena Eggers¹, Antonia Thielecke¹, Mar Fernandez Mendez^1,2

1: Alfred Wegener Institute, Germany; 2: University of Bremen, Germany

SoraCT: Unconditioned 3D CT Synthesis via Video Diffusion Transformers

Jiayi Wang¹, Hadrien Reynaud¹, Bernhard Kainz^1,2

1: Friedrich-Alexander University Erlangen-Nürnberg, Germany; 2: Department of Computing, Imperial College London, UK

A 3D Foundation Model for Generalizable Biological Structure Segmentation in Tissue Clearing Images

Qi Liu^1,3,4,5, Ying Chen^1,3,4,5, Markus Elsner^1,6, Ali Ertürk^1,2,4,6,7

1: Institute for Stroke and Dementia Research, Klinikum der Universität München, Ludwig-Maximilians University Munich, Munich, Germany; 2: Institute for Intelligent Biotechnologies (iBIO), Helmholtz Center Munich, Neuherberg, Germany; 3: Faculty of Medicine, Ludwig-Maximilians University Munich, Munich, Germany; 4: Munich Cluster for Systems Neurology (SyNergy), Munich, Germany; 5: Munich Medical Research School (MMRS), Munich, Germany; 6: Deep Piction GmbH, Munich, Germany; 7: School of Medicine, Koç University, İstanbul, Turkey

Factor dependencies and their associations with covariates in multimodal factor analysis

Małgorzata Łazęcka¹, Ewa Szczurek^2,1

1: University of Warsaw, Poland; 2: Helmholtz Center Munich, Germany

The Age of Artificial Work - What happens when humans are no longer the only ones working?

Guido Mathews

Entrepreneur, Germany

Developing Agile and Collaborative Platforms for AI Governance in Research Institutions

Veit Schwab¹, Manuel Hoffmann²

1: Max Planck Institute of Psychiatry, Germany; 2: Max Planck Information and Technology (MaxIT), Germany

Data2AI Compass

Sonja Hänzelmann, Luca Greiner

Alfred Wegener Institute for Polar and Marine Sciences, Germany

The Future of Universal Inverse Problems Solvers

Dawit Abiy Hailu

Helmholtz Zentrum Hereon, Germany

Large Vision-Language Models for Earth Observation Applications

Luca Chiarabini

DLR, Germany

AI Engineering for Safety-Critical Applications

Yannick Kees

German Aerospace Center (DLR), Germany

After Generative AI: Is Physical Intelligence the Next Wave for Science?

Celso Rêgo

Karlsruhe Institute of Technology, Germany

Can AI Become a Responsible Teaching Partner? Using, Improving and Creating Educational Resources with Generative AI

Philip Müller¹, Donatella Cea², Sebastian Starke¹, Gunjan Joshi¹

1: Helmholtz Zentrum Dresden-Rossendorf (HZDR), Germany; 2: Helmholtz AI, Helmholtz Centre Munich, Germany

Inclusion and Diversity in AI and Biomedical Research

Beatrice Guastella^1,2, Joel Valdivia Ortega^1,3,4, Qian Xu^1,5, Dennis Ostländer^1,6, Luisa Hernández Götz^1,7

1: Helmholtz Munich, Germany; 2: LMU Klinikum, Germany; 3: Helmholtz Pioneer Campus, Germany; 4: Helmholtz AI, Germany; 5: Technical University of Munich, Germany; 6: University of Regensburg, Germany; 7: University of Edinburgh, United Kingdom

The Architect and the Engine: Catching "Invisible Failures" in AI-Generated Code

Nico Schuster

Centre de Physique des Particules de Marseille, France

How To Provide (Unified) Access To AI Models?

Jonas Hagenberg

Helmholtz Zentrum München, Germany

Managing AI/ML lifecycle: tools and needs

Lisana Berberi, Valentin Kozlov, Paul Kieckhefen

Karlsruhe Institute of Technology (KIT), Germany

Federated Learning: Fine-Tuning Large AI Models Across Institutions - Without Sharing Data

Leonhard Duda, Khadijeh Alibabaei

KIT, Germany

To what extend Can AI Replace Numerical Simulations in Engineering? Opportunities, Challenges, and Limits

Tarek Hussein

Universität Augsburg, Germany

Scholarly Publishing in the Age of AI: Tips, the Editor's Black Box, and AI in Publishing

Hashini Thirimanne

Wiley, United Kingdom

Can Urban AI Make Cities More Human?

Narendran Sobanapuram Muruganandam, Erika Von Schneidemesser

Research Institute for Sustainability, Germany

Beyond Prediction: Can AI Make Treatment Decisions?

Pamela Solano

Faculty of Computer Science and Data Science. Regensburg University, Germany

Interpretable Representations for Hematology

Muhammed Furkan Dasdelen¹, Hyesu Lim¹, Paul Pommer¹, Michele Buck², Steffen Schneider¹, Carsten Marr¹

1: Helmholtz Munich, Germany; 2: Medical Department for Hematology and Oncology, Technical University Munich

Analysis of multi-omics data using graph neural networks identifies novel schizophrenia-associated genes suitable as drug targets

Svitlana Oleshko^1,2, Samuele Firmani¹, Jamie Robinson², Julien Hernandez-Lallement², Sergio Picart-Armada², Matthias Heinig¹, Till F.M. Andlauer^2,3, Annalisa Marsico²

1: Helmholtz Munich, Germany; 2: Boehringer Ingelheim Pharma, Germany; 3: Technical University of Munich, Germany

MLflow pilot service for Helmholtz researchers

Lisana Berberi, Christophe Laures, Khadijeh Alibabaei, Valentin Kozlov, Achim Streit

Karlsruhe Institute of Technology (KIT), Germany

Towards Robust Foundation Models for Digital Pathology

Jonah Kömen^1,2, Edwin de Jong³, Julius Hense^1,2, Hannah Marienwald^1,2, Jonas Dippel^1,2,3, Philip Naumann^1,2, Eric Marcus⁴, Lukas Ruff³, Maximilian Alber^3,7, Jonas Teuwen⁴, Frederick Klauschen^1,5,6,7, Klaus-Robert Müller^1,2,8,9

1: Berlin Institute for the Foundations of Learning and Data (BIFOLD), Berlin, Germany; 2: Machine Learning Group, Technische Universität Berlin, Berlin, Germany; 3: Aignostics, Berlin, Germany; 4: The Netherlands Cancer Institute Amsterdam (NKI), Antoni van Leeuwenhoek Hospital (AvL), Amsterdam, Netherlands; 5: Institute of Pathology, Ludwig Maximilian University, Munich, Germany; 6: German Cancer Research Center, Heidelberg, and German Cancer Consortium, Munich, Germany; 7: Institute of Pathology, Charité Universitätsmedizin, Berlin, Germany; 8: Department of Artificial Intelligence, Korea University, Seoul, Korea; 9: Max-Planck Institute for Informatics, Saarbrücken, Germany

Sailing Past Syntax: A Human-in-the-Loop Framework for Safe Generative AI in Science

Nico Schuster

Centre de Physique des Particules de Marseille, France

The Helmholtz Model Zoo: Enabling AI Model Sharing and Inference in the Helmholtz Cloud

Hans Werners, Engin Eren, Philipp Heuser

Deutsches Elektronen-Synchroton DESY, Germany

Improving Reliability of LLM-Based Robotic Task Planning Through Domain Adaptation and Benchmarking

Kanak Pandit, Ananya Kulkarni, Muhammad Saeed

ARENA2036, Germany

Uncertainty-Guided Generation of Dark-Field Radiographs

Lina Felsner^1,2,3, Henriette Bast^4,5,6, Tina Dorosti^4,5,6, Florian Schaff^4,5,6, Franz Pfeiffer^4,5,6,7, Daniela Pfeiffer^5,6,7, Julia Schnabel^1,2,3,8

1: School of Computation, Information and Technology, Technical University of Munich, 85748 Garching, Germany; 2: Munich Center for Machine Learning (MCML), Munich, Germany; 3: Institute of Machine Learning in Biomedical Imaging, Helmholtz Munich, 85764 Neuherberg, Germany; 4: Department of Physics, School of Natural Sciences, Technical University of Munich, 85748 Garching, Germany; 5: Munich Institute of Biomedical Engineering, Technical University of Munich, 85748 Garching, Germany; 6: Institute for Diagnostic and Interventional Radiology, School of Medicine and Health, TUM University Hospital Klinikum rechts der Isar, Technical University of Munich, 81675 Munich, Germany; 7: Institute for Advanced Study, Technical University of Munich, 85748 Garching, Germany; 8: School of Biomedical Engineering & Imaging Sciences, King’s College, London, UK

SoraCT: Unconditioned 3D CT Synthesis via Video Diffusion Transformers

Jiayi Wang¹, Hadrien Reynaud¹, Bernhard Kainz^1,2

1: Friedrich-Alexander University Erlangen-Nürnberg, Germany; 2: Department of Computing, Imperial College London, UK

A 3D Foundation Model for Generalizable Biological Structure Segmentation in Tissue Clearing Images (DISCO-CAR: whole-body mapping of engineered cells and diseases at single-cell resolution)

Qi Liu^1,3,4,5, Jennifer Modamio², Ying Chen^1,3,4,5, Doris Kaltenecker², Karoline Kadletz², Izabela Horvath^2,4,6,8, Luciano Hoeher², Eva Küsters², David-Paul Minde², Chenchen Pan^1,2, Hongcheng Mai^1,3,4,6, Dirk H Busch^9,10,11, Markus Elsner^1,6, Ali Ertürk^1,2,4,6,7

1: Institute for Stroke and Dementia Research, Klinikum der Universität München, Ludwig-Maximilians University Munich, Munich, Germany; 2: Institute for Intelligent Biotechnologies (iBIO), Helmholtz Center Munich, Neuherberg, Germany; 3: Faculty of Medicine, Ludwig-Maximilians University Munich, Munich, Germany; 4: Munich Cluster for Systems Neurology (SyNergy), Munich, Germany; 5: Munich Medical Research School (MMRS), Munich, Germany; 6: Deep Piction GmbH, Munich, Germany; 7: School of Medicine, Koç University, İstanbul, Turkey; 8: TUM School of Computation, Information and Technology, Technical University of Munich, Munich, Germany; 9: German Center for Infection Research (DZIF), Partner Site Munich, Munich, Germany; 10: Institute for Medical Microbiology, Immunology and Hygiene, Technische Universität München (TUM), Munich, Germany; 11: Focus Group 'Clinical Cell Processing and Purification', Institute for Advanced Study, TUM, Munich, Germany

M²S³OM-graph: A Hybrid Deterministic-LLM Pipeline for Automated Metadata Crosswalk Extraction and Graph-Grounded Interoperability

Santiago Casas Castro

Deutsches Zentrum für Luft- und Raumfahrt e. V. (DLR), Germany

EO Foundation Models for Pre-Training Dataset Uncertainty Analysis

Stefan Dan, Jakob Gawlikowski

German Aerospace Center, Remote Sensing Technology Institute, Germany

FrustrAI-Seq: Scaling Local Energetic Frustration to the Protein Sequence Space

Jan-Philipp Leusch^1,2, Miriam Poley-Gil³, Miguel Fernandez-Martin³, Nicola Bordin⁴, Burkhard Rost², R. Gonzalo Parra³, Michael Heinzinger^1,2

1: Helmholtz Munich, Germany; 2: Technical University of Munich, Germany; 3: Barcelona Supercomputing Center, Spain; 4: University College London, United Kingdom

muBRAND: Multi-Branch Diffusion for Flexible Generation and Inpainting of Image Series

Jan-Oliver Kropp^1,2,3, Katrin Amunts^1,3, Timo Dickscheid^1,2,4, Christian Schiffer^1,2

1: Forschungszentrum Jülich, Germany; 2: Helmholtz AI, Jülich, Germany; 3: C. und O. Vogt Insitut, Universitätsklinikum Düsseldorf, Düsseldorf, Germany; 4: Institute of Computational Visualistics, Koblenz University, Koblenz, Germany

Multimodal Analysis of Organoid Network Dynamics Using Multielectrode Arrays and Calcium Imaging

Katrina Ager, Silviu Vasile-Bodea, Lucas Mair, Gil Westmeyer, Steffen Schneider

Helmholtz Munich, Germany

Towards Multimodal Inference of Electron Bunch Structures using Gradient-Based Reconstruction with Differentiable Detector Models

Ritz Ann Aguilar¹, Maxwell LaBerge¹, Alexander Debus¹, Andreas Doepp², Zewu Bi^2,3, Michael Bussmann^1,4, Arie Irman¹, Ulrich Schramm^1,5, Jeffrey Kelling^1,6

1: Helmholtz-Zentrum Dresden-Rossendorf, Germany; 2: Ludwig-Maximilians-Universität München, Germany; 3: Shanghai Jiao Tong University, China; 4: Center for Advanced Systems Understanding, Germany; 5: Technische Universität Dresden, Germany; 6: Technische Universität Chemitz, Germany

NURBSeg: bridging the gap between medical imaging and CAD via end-to-end parametric segmentation

Hakan Akdag¹, Philipp Rosauer¹, Sebastian Ziegler², Katharina Pröll¹, Kris Dreher², Alexander Rüttgers¹

1: German Aerospace Center (DLR), Germany; 2: German Cancer Research Center (DKFZ), Germany

From prediction to generation of antigen-specific, full-length T-cell receptors

Yang An^1,2, Felix Drost^1,2, Adrian Straub², Kersten Heyer², Annalisa Marsico¹, Dirk H. Busch², Benjamin Schubert¹

1: Helmholtz Munich, Germany; 2: Technical University of Munich, Germany

Provenance-Aware Knowledge Graphs for Reusable PEFC and PEWE Experiments

Ezgi Arat^1,2,3, Andreas Glüsen¹, Martin Florian Seidler¹, Mohammad Javad Eslamibidgoli², Michael Eikerling^2,3, Kourosh Malek², Paul Kieckhefen^4,5

1: Institute of Energy Technologies (IET-4), Electrochemical Process Engineering, Forschungszentrum Jülich GmbH, 52425 Jülich, Germany; 2: Theory and Computation of Energy Materials (IET-3), Institute of Energy Technologies, Forschungszentrum Jülich GmbH, 52425 Jülich, Germany; 3: Chair for Theory and Computation of Energy Materials, Faculty of Georesources and Materials Engineering, RWTH Aachen University, 52062 Aachen, Germany; 4: Data Analytics, Access and Applications, Scientific Computing Center (SCC), Karlsruher Institut für Technologie, 76344 Eggenstein-Leopoldshafen, Germany; 5: Helmholtz AI, Karlsruher Institut für Technologie, 76344 Eggenstein-Leopoldshafen, Germany

Deep Learning Modeling of RNA Structure Perturbations Induced by Small Molecules

Tobias Bernecker^1,2, Lambert Moyon¹, Danny Incarnato^3,4, Annalisa Marsico^1,4

1: Computational Health Center, Helmholtz Munich, Munich, Germany; 2: School of Computation, Information and Technology, Technical University of Munich, Munich, Germany; 3: Department of Molecular Genetics, Groningen Biomolecular Sciences and Biotechnology Institute (GBB), University of Groningen, Groningen, The Netherlands; 4: Equal contributions

T-time: Multimodal Manifold Learning for Clonally Constrained Trajectory Inference

Irene Bonafonte Pardàs¹, Myriam Lizotte^2,3, Guy Wolf^2,3, Benjamin Schubert¹

1: Helmholtz Munich, Germany; 2: Mila – Quebec AI Institute, Montréal, Canada; 3: Department of Mathematics and Statistics, Université de Montréal

End-To-End Protein Engineering With Protein Language Models

Maurice Brenner^1,2, Helene Wetekam^1,2, Alexander Plaikner², Manuel Ploner², Robert Schmirler², Burkhard Rost², Michael Heinzinger^1,2

1: Institute of Computational Biology, Computational Health Center, Helmholtz Munich, Germany; 2: Department of Bioinformatics and Computational Biology , Technical University of Munich, Germany

AI-assisted Labeling and its Pitfalls: A Case Study in Electron Microscopy Segmentation

Christina Bukas¹, Jonas Engler², Helena Pelin¹, Jonas Hagenberg¹, Haider Khan¹, Francesco Campi¹, Carola Eberhagen³, Hans Zischka^2,3, Marie Piraud¹

1: Helmholtz AI, Helmholtz Center Munich, Germany; 2: Institute of Toxicology and Environmental Hygiene, TUM School of Medicine and Health, Technical University of Munich, Germany; 3: Institute of Molecular Toxicology and Pharmacology, Helmholtz Center Munich, Germany

Spatial and Temporal Evaluation of Extreme Precipitation in ML-Based Global Weather Forecasts

Sebastian Buschow, Wael Almikaeel

Forschungszentrum Jülich, Germany

Recurrent Neural Networks and Tensor Ring Decomposition

Saptarshi Chakrabarti, Nick Vannieuwenhoven

KU Leuven, Belgium

AI and Hierarchical Clustering Techniques for Accurate Patient Stratification

Juan G. Diaz Ochoa², Miroslav Puskaric³, Natalie Layer⁴, Antje Jensch⁴, Markus Knott⁴, Alexander Krohn^1,5

1: Hochschule München, Campus for Health and Engineering, Munich, Germany; 2: PerMediQ, Stuttgart, Germany; 3: University of Stuttgart, HLRS - High Performance Computing Center, Stuttgart, Germany; 4: Klinikum Stuttgart, Stuttgart Cancer Center - Tumorzentrum Eva Mayr-Stihl DE, Stuttgart, Germany; 5: University Hospital Regensburg, Department of Internal Medicine I, Regensburg, Germany

SciTracer - the reliable and logically provable autonomous scientific discovery.

Leonid Didukh

Unaffiliated, Ukraine

Benchmarking for Microscopic Biodiversity Screening (AIMBIS-project)

Susanne Dunker^1,2, Kathleen Stoof-Leichsenring³, Mar Fernández-Méndez⁴, Riccardo Massei^1,2, Jan Bumberger^1,2

1: Helmholtz-Centre for Environmental Research GmbH (UFZ), Germany; 2: German Centre for integrative Biodiversity Research e.V. Halle-Jena-Leipzig (iDiv), Germany; 3: Helmholtz-Centre for Polar and Marine Research – AWI, Department/Section Polar Terrestrial Environmental Systems; 4: Helmholtz-Centre for Polar and Marine Research – AWI, Department/Section Polar Biological Oceanography

MolecuLA: Learning Molecules As A Language For Generative And Interpretable Chemistry

Zakaria Elabid¹, Attila Cangi¹, Jan Andrzejewski^1,2, Bartosz Brzoza^1,2

1: CASUS/Helmholtz-Zentrum Dresden-Rossendorf, Germany; 2: University of Wrocław, Poland

Construct-valid and adaptive benchmarks for single-cell RNA-seq via a hierarchical, streaming simulator

Philipp Flotho^1,2, Matthias Flotho^1,2, Andreas Keller^1,2

1: Chair for Clinical Bioinformatics, Saarland University, Germany; 2: Helmholtz Institute for Pharmaceutical Research Saarland

From T-FAKE to TFAN: Reusable Thermal Face Alignment for AI in Scientific Imaging

Philipp Flotho^1,4,5, Moritz Piening^2,5, Anna Kukleva³, Gabriele Steidl²

1: Chair for Clinical Bioinformatics, Saarland University, Germany; 2: Technical University of Berlin, Germany; 3: MPI for Informatics, SIC, Germany; 4: Helmholtz Institute for Pharmaceutical Research, Germany; 5: Equal author contribution

From Feature Learning to Neural Collapse: A Data-Dependent Kernel View

Pankhil Gawade^1,2, Kirsten Fischer¹, Ewa Szczurek^1,3

1: Helmholtz Munich, Germany; 2: Technical University of Munich, TUM School of Computation, Information and Technology; 3: Faculty of Mathematics, Informatics and Mechanics, University of Warsaw

Do world models model the world?

Rodrigo Gonzalez Laiz

Helmholtz Munich, Germany

Editing the Foundations: Task Arithmetic for 3D Medical Segmentation Foundation Models

Cristina González^1,2, Laura Daza^1,2, Julia Schnabel^1,2,3

1: Helmholtz Munich, Germany; 2: Technical University of Munich, Germany; 3: King's College London, UK

Spatial transcriptomics-informed inference of cell types and states from nuclear-stained whole slide images.

Beatrice Guastella^1,2, Ziqing Zhao^1,2, Martin Dichgans^2,4, Mikael Simons^2,3,4,5, Hannah Spitzer^1,2,4

1: Institute of Computational Biology (ICB), Helmholtz Zentrum München, Munich, Germany; 2: Institute for Stroke and Dementia Research (ISD), Klinikum der Universität München, Munich, Germany; 3: German Center for Neurodegenerative Diseases (DZNE), Munich, Germany; 4: Munich Cluster of Systems Neurology (SyNergy), Munich, Germany; 5: Institute of Neuronal Cell Biology, Technical University Munich, Munich, Germany

Uncertainty Quantification for Medical Image Segmentation

Maleeha Hassan, Till Korten, Ulrik Günther, Peter Steinbach

HZDR, Germany

Fast and Few-Shot: Equipping Scientific LLM Agents with Meta-Learned Surrogates

Hamed Hemati¹, Stefan Sandfeld^1,2

1: Institute for Advanced Simulations – Materials Data Science and Informatics (IAS-9), Forschungszentrum Jülich GmbH, 52425 Jülich, Germany; 2: Chair of Materials Data Science and Materials Informatics, Faculty 5 – Georesources and Materials Engineering, RWTH Aachen University, 52056 Aachen, Germany

scooby: modeling multimodal genomic profiles from DNA sequence at single-cell resolution.

Johannes Hingerl^1,3, Laura D. Martens^1,2, Aleksandr Neverov¹, Mariia Minaeva², Nils F. Wagner¹, Fabian J. Theis^1,2, Julien Gagneur^1,2,3,4

1: Technical University of Munich, School of Computation, Information and Technology, Munich, Germany; 2: Helmholtz Center Munich, Computational Health Center, Neuherberg, Germany; 3: Technical University of Munich, Munich Center for Machine Learning, Munich, Germany; 4: Technical University of Munich, Institute of Human Genetics, School of Medicine, Munich, Germany

Geospatial Causal Inference for Drought Driver Analysis

Vytautas Jancauskas, Samuel Garske, Daniela Espinoza-Molina

DLR, Germany

When Protein Dynamics Matter: Integrating Molecular Dynamics into Protein Foundation Models

Reem Aboelsoud, Kasravi Javad, Stefan Kesselheim, Alina Bazarova

jülich forschungszentrum, Germany

Toward Automated MRI-Based Screening of Cortical Superficial Siderosis Using Multimodal 3D nnU-Net

Wasim Aftab¹, Anna Dewenter¹, Maria Kaffe¹, Anna Kopczak¹, Martin Dichgans¹, Hannah Spitzer^1,2

1: Institute for Stroke and Dementia Research(ISD), University Hospital, LMU Munich,Munich, Bavaria, Germany; 2: Institute of Computational Biology (ICB), Helmholtz Munich, Neuherberg, Germany

Cross-View Latent Integration via Nonparametric Gamma Shrinkage Factor Analysis

Hassan Akell¹, Małgorzata Łazęcka², Dinesh A Haridoss¹, Miriam Urban³, Eike Staub³, Ewa Szczurek^1,2

1: Institute of AI for Health, Helmholtz Munich; 2: Faculty of Mathematics, Informatics and Mechanics, University of Warsaw; 3: Oncology Data Science, Merck Healthcare KGaA

Physics-Guided Diffusion for Neonatal Ultra-Low-Field MRI enhancement

Malek Al Abed^1,3, Sebiha Demir², Anne Groteklaes², Elodie Germani⁴, Shahrooz Faghihroohi³, Hemmen Sabir², Shadi Albarqouni^1,3

1: University of Bonn, University Hospital Bonn, Clinic for Diagnostic and Interventional Radiology, Germany; 2: University of Bonn, University Hospital Bonn, Department of Experimental Neonatology, Germany; 3: Technical University of Munich, School of Computation, Information and Technology, Germany; 4: Laboratoire Traitement du Signal et de l’Image, Inserm UMR 1099, Universit´ e de Rennes, France

Atlas 2 - Foundation Models for Clinical Deployment

Maximilian Alber^1,13, Timo Milbich¹, Alexandra Carpen-Amarie¹, Stephan Tietz¹, Jonas Dippel^1,7,8, Lukas Muttenthaler^1,15,16, Beatriz Perez Cancer¹, Alessandro Benetti¹, Panos Korfiatis³, Elias Eulig¹, Jérôme Lüscher¹, Jiasen Wu¹, Sayed Abid Hashimi¹, Gabriel Dernbach^1,8,13, Simon Schallenberg¹³, Neelay Shah¹, Moritz Krügener¹, Aniruddh Jammoria¹, Jake Matras⁵, Patrick Duffy⁶, Matt Redlon⁴, Philipp Jurmeister^11,12, David Horst^11,13, Lukas Ruff¹, Klaus-Robert Müller^7,8,9,10, Frederick Klauschen^12,13,14, Andrew Norgan²

1: Aignostics, Germany; 2: Department of Laboratory Medicine and Pathology, Mayo Clinic, Rochester, MN, US; 3: Department of Radiology, Mayo Clinic, Rochester MN, US; 4: Department of Information Technology, Mayo Clinic, Rochester MN, US; 5: Mayo Clinic, Rochester MN, US; 6: Digital Pathology, Mayo Clinic, Rochester MN, US; 7: Machine Learning Group, Technische Universität Berlin, Germany; 8: BIFOLD – Berlin Institute for the Foundations of Learning and Data, Germany; 9: Department of Artificial Intelligence, Korea University, Republic of Korea; 10: Max-Planck Institute for Informatics, Germany; 11: German Cancer Research Center (DKFZ) & German Cancer Consortium (DKTK), Berlin & Munich Partner Sites, Germany; 12: Institute of Pathology, Ludwig-Maximilians-Universität München, Germany; 13: Institute of Pathology, Charité – Universitätsmedizin Berlin, Germany; 14: Bavarian Cancer Research Center (BZKF), Germany; 15: Helmholtz Munich, Germany; 16: Technical University Munich, Germany

EOSC-ARENA: Towards an Open and Ethical GenAI Infrastructure for European Research

Khadijeh Alibabaei¹, Lisana Berberi¹, Valentin Kozlov¹, Alexandra Delipalta², Álvaro López García³, Germán Moltó⁴, Fotis Psomopoulos⁵, Viet Tran⁶, Marcin Plociennik⁷

1: Scientific Computing Center (SCC), Karlsruhe Institute of Technology (KIT), Eggenstein-Leopoldshafen, 76344,Germany; 2: Research Data Alliance Association - RAD, Belgium; 3: Instituto de Física de Cantabria (IFCA), CSIC-UC, Avda. los Castros s/n, Santander, 39006, Cantabria, Spain; 4: Instituto de Instrumentación para Imagen Molecular (I3M), Centro Mixto CSIC — Universitat Politècnica de València, Camino de Vera s/n, 46022 Valencia, España; 5: Ethniko Kentro Erevanas Kai Technologikis Anaptyxis - CERTH, Greece; 6: nstitute of Informatics, Slovak Academy of Sciences (IISAS), Dúbravská cesta 9, Bratislava, 84507, Slovakia; 7: Instytut Chemii Bioorganicznej Polskiej Akademii Nauk - PSNC, Poland

Multimodal graph-based fusion learning for Alzheimer’s disease prediction

Shubhi Ambast^1,2, Karola Mai^1,2, Charles Mwangi Kaumbutha^1,2, Jannis Spintge², Caterina Carraro^2,4, Marc Beyer^2,3, Matthias Becker^1,2

1: Modular High-Performance Computing and Artificial Intelligence, German Center for Neurodegenerative Diseases (DZNE), Bonn, North Rhine-Westphalia, Germany; 2: Systems Medicine, German Center for Neurodegenerative Diseases (DZNE), Bonn, North Rhine-Westphalia, Germany; 3: PRECISE Platform for Genomics and Epigenomics at DZNE and University of Bonn, Bonn, North Rhine-Westphalia, Germany; 4: Life and Medical Sciences (LIMES) Institute, Genomics & Immunoregulation, Bonn, North Rhine-Westphalia, Germany

Can Large Language Models Grade Short Free-Text Exam Answers? An Evaluation on Real German Student Data

Zeynep Arslan, Lennart Bressel, Fritz Niesel

Forschungszentrum Jülich, Germany

Pseudo-Anomaly Augmented One-Class Autoencoder and Supervised Deep Learning for Murine ECG Anomaly Detection

Marcella Astrid¹, Ewan Selkirk², Christopher S. Ward³, Donatella Cea¹, Christina Bukas¹, Elida Schneltzer⁴, Marie Piraud¹, Nadine Spielmann⁴

1: Helmholtz AI, Helmholtz Center Munich, Germany; 2: Mary Lyon Centre at MRC Harwell, Harwell Campus, United Kingdom; 3: Integrative Physiology/Advanced Technology Cores, Baylor College of Medicine, Texas, United States; 4: Institute of Experimental Genetics, Helmholtz Center Munich, German Research Center for Environmental Health, Germany

A Physics-Based Machine Learning Workflow for Urban-Scale Residential Heat-Demand Estimation

Nazanin Bagherinejad, V Mithlesh Kumar, Julia Kowalski

Chair of Methods for Model-based Development in Computational Engineering, RWTH Aachen University, Germany

scDigital-Karyotypes: a genomic representation to learn shared variants

Suharto Banerjee¹, Elisabeth Georgii², Lisa Barros de Andrade e Sousa², Mai Abdelhalim Mohamed³, Karen Lopez-Cayuqueo³, Kai M. Schmidt-Ott³, Ashley D. Sanders^1,4

1: Max Delbrück Center for Molecular Medicine, Berlin, Germany; 2: Helmholtz AI, Helmholtz Munich, Germany; 3: Department of Nephrology and Hypertension, Hannover Medical School, Hannover, Germany; 4: Charité – Universitätsmedizin Berlin, Germany

Semi-Supervised Federated Learning for Medical Image Classification

Ayse Edanur Baser

Ozyegin University, Turkey (Türkiye)

ECG Template-Based Deep Learning for Amplified P-wave Duration Prediction

Silvia Becker^1,2, Ajay Krishna², Heiko Lehrmann², Amir Jadidi³, Till Keller^2,4, Dirk Westermann², Thomas Arentz², Axel Loewe¹, Martin Eichenlaub²

1: Karlsruhe Institute of Technology, Germany; 2: University Heart Center Freiburg - Bad Krozingen, Germany; 3: Lucerne Cantonal Hospital, Switzerland; 4: Justus Liebig University Giessen, Germany

Reduction of Outflow Boundary Influence on Aerodynamic Performance Using Neural Networks

Mario Christopher Bedrunka^1,2,3,4, Dirk Reith^4,5, Holger Foysi³, Łukasz Łaniewski-Wołłk⁶, Travis Mitchell²

1: Jülich Supercomputing Centre (JSC), Forschungszentrum Jülich, Wilhelm-Johnen-Straße 52428 Jülich, Germany; 2: School of Mechanical and Mining Engineering, The University of Queensland, St Lucia 4072, Queensland, Australia; 3: Chair of Fluid Mechanics, University of Siegen, Paul-Bonatz-Straße 9-11, Siegen-Weidenau 57076, Germany; 4: Institute of Technology, Resource and Energy-efficient Engineering (TREE), Bonn-Rhein-Sieg University of Applied Sciences, Grantham-Allee 20, Sankt Augustin 53757, Germany; 5: Fraunhofer Institute for Algorithms and Scientific Computing (SCAI), Schloss Birlinghoven, Sankt Augustin 53754, Germany; 6: Institute of Aeronautics and Applied Mechanics, Warsaw University of Technology, Nowowiejska 24, Warszawa 00-665, Poland

Rolv.io - Multi-Agent AI For Healthcare And Research

Alexander Blume¹, Deniz Aydoğdu^1,2, Yoga Mithra Senthil Kumar¹, Nelli Khachatryan¹, Ahmed Tijani Akinfalabi¹, Arash Sorooshmehr¹, Ricardo Wurmus¹, Altuna Akalin¹

1: Max Delbrück Center (MDC), Germany; 2: Hacettepe University, Turkey

Understanding Green Technology Adoption in Germany: A Cluster-Based Analysis of Household-Scale Photovoltaics and Battery Electric Vehicles

Philipp C. Böttcher¹, Leo Kaindl², Lisa Barros de Andrade e Sousa², Marie Piraud², Sandra Venghaus^3,4, Dirk Witthaut^1,5

1: Institute of Climate and Energy Systems: Energy Systems Engineering (ICE-1), Forschungszentrum Jülich, Germany; 2: Helmholtz AI, Helmholtz Munich, Germany; 3: Institute of Climate and Energy Systems: Jülich Systems Analysis (ICE-2), Forschungszentrum Jülich, Germany; 4: School of Business and Economics, RWTH Aachen University, Germany; 5: Institute for Theoretical Physics, University of Cologne, Germany

From Learned Representation To Localization: Towards Semantic Integration Of Microscopic Data In The Human Brain

Zeynep Boztoprak^1,2, Katrin Amunts^1,4, Markus Axer^1,5, Timo Dickscheid^1,2,3, Christian Schiffer^1,2

1: Institute of Neuroscience and Medicine (INM-1), Research Centre Jülich, Germany; 2: Helmholtz AI, Research Centre Jülich, Germany; 3: Computer Vision, Institute for Computational Visualistics, University of Koblenz, Germany; 4: C. & O. Vogt Institute of Brain Research, University Hospital Düsseldorf, Germany; 5: Department of Physics, University of Wuppertal, Germany

AI-assisted Classification of Polar Phytoplankton’s Functional Biodiversity using Multispectral Imaging Flow Cytometry

Natasha Bryan¹, Anna Naberhaus¹, Sarah Lena Eggers¹, Antonia Thielecke¹, Mar Fernandez Mendez^1,2

1: Alfred Wegener Institute, Germany; 2: University of Bremen, Germany

LEVERAGING MULTI-RATER ANNOTATIONS TO CALIBRATE OBJECT DETECTORS IN MICROSOPY IMAGING

Francesco Campi, Marie Piraud

Helmholtz AI, Helmholtz Zentrum Muenchen, Germany

CytoDiff: AI-Driven Cytomorphology Image Synthesis for Medical Diagnostics

Jan Carreras Boada, Rao Muhammad Umer, Carsten Marr

Helmholtz Munich, Germany

DeepElbe: Multi-Horizon Oxygen Forecasting and Hypoxia Early Warning Using Statistical and Deep Learning Methods

Danu Caus, Ovidio Fernando Garcia Oliva, Carsten Lemmen, Tobias Weigel

Helmholtz-Zentrum Hereon, Germany

Cancer-label-free Tumor Segmentation in Whole-mouse Light Sheet Microscopy Images Using Vision-language Models

Ying Chen^1,3,4,5, Eva Küsters^2,6, Jie Zhang^1,3,4,5, Markus Elsner^2,7, Ali Ertürk^1,2,4,7,8

1: Institute for Stroke and Dementia Research, Klinikum der Universität München, Ludwig-Maximilians University Munich, Munich, Germany; 2: Institute for Intelligent Biotechnologies (iBIO), Helmholtz Center Munich, Neuherberg, Germany; 3: Faculty of Medicine, Ludwig-Maximilians University Munich, Munich, Germany; 4: Munich Cluster for Systems Neurology (SyNergy), Munich, Germany; 5: Munich Medical Research School (MMRS), Munich, Germany; 6: Graduate School of Neuroscience (GSN), Munich, Germany; 7: Deep Piction GmbH, Munich, Germany; 8: School of Medicine, Koç University, İstanbul, Turkey

Physics-Informed Reconstruction of Electron-Bunch Shapes from CTR Spectra using an Untrained Neural Network

Fabia Dietrich^1,2, Ritz Ann Aguilar¹, Alexander Debus¹, Arie Irman¹, Maxwell LaBerge¹, Richard Pausch¹, Jessica Tiebel^1,2, Nico Wrobel^1,2, Ulrich Schramm^1,2, Michael Bussmann^1,3, Jeffrey Kelling^1,4

1: Helmholtz-Zentrum Dresden-Rossendorf, Germany; 2: Technische Universität Dresden, Germany; 3: Center for Advanced Systems Understanding, Germany; 4: Technische Universität Chemnitz, Germany

Modeling Long Contexts with Hierarchical Compression Transformers

Eske Ewert, Mario Christopher Bedrunka, Stefan Kesselheim

Forschungszentrum Jülich, Germany

Towards high resolution air quality forecasts for Europe using WGAN-based statistical downscaling with EURAD-IM data

Philipp Franke¹, Sindhu Vasireddy¹, Michael Langguth², Anne Caroline Lange¹, Martin Schultz¹, Anke C. Nölscher¹

1: Forschungszentrum Jülich, Germany; 2: Quarda Energy GmbH, Germany

Data-Driven Identification of Possible Transmission Sites for Carbapenem-Resistant Gram-Negative Bacteria

Elisabeth Georgii¹, Petra Heinmüller², Philipp Thomas², Lisa Barros de Andrade e Sousa¹, Theresa Willem¹, Ivo Foppa², Marie Piraud¹, Anja M. Hauri³

1: Helmholtz AI, Helmholtz Munich, Germany; 2: Hessian State Office for Health and Care (HLfGP), Germany; 3: Helmholtz Centre for Infection Research, Germany

Topological Data Analysis of Single-Cell Multimodal Data

Isabel Giray^1,2, Gabriele Malagoli^1,3, Katharina Schmid¹, Maria Colomé-Tatché^1,3,4

1: Biomedical Center (BMC), Physiological Chemistry, Faculty of Medicine, LMU Munich, Munich, Planegg-Martinsried, Germany; 2: TUM, Germany; 3: Institute of Computational Biology, Computational Health Center, Helmholtz Zentrum München, German Research Center for Environmental Health, Neuherberg, Germany; 4: Hospital del Mar Research Institute (HMRIB), Barcelona, Spain

Are Soft Labels Worth the Effort?

Blagovest Gospodinov, Vincent Fortuin

Helmholtz, Germany

Building a Local AI Agent Assisting with Sensor Observation to Archives Framework (O2A)

Luca Greiner¹, Stefan Pinkernell¹, Roland Koppe², Sonja Hänzelmann¹

1: Alfred-Wegener-Institute for Polar and Marine Research, Data Science, Bremerhaven, Bremen, Germany; 2: Alfred-Wegener-Institute for Polar and Marine Research, Software Engineering, Bremerhaven, Bremen, Germany

SEA2LAND Navigator-GPT

Harsh Grover¹, Tobias Weigel¹, Manuel Frias², David Cabana³, Louis Celliers³, Anu Printsmann⁴

1: Hereon, Germany; 2: HELCOM; 3: GERICS; 4: Tallinn University

Messanger RNA Untranslated Region Design through Transformers

Lorenzo Guerci¹, Ewa Szczurek^1,2

1: Helmholtz Munich, Institute of Computational Biology, Germany; 2: Faculty of Mathematics, Informatics and Mechanics, University of Warsaw

Semantic Equilibrium Decoding for Open-Ended Medical Visual Question Answering

Luca Hagen¹, Johanna Paula Müller¹, Weitong Zhang², Mengyun Qaio³, Bernhard Kainz^1,2

1: FAU Erlangen-Nürnberg, Germany; 2: Imperial College London, UK; 3: UCL, UK

Single Cell PBMC Analysis Shows Cell Type Specific Changes On Inflammation In Both Depressed And Non-Depressed Individuals

Jonas Hagenberg^1,2, Janine Knauer-Arloth²

1: Helmholtz AI, Helmholtz Zentrum München, Germany; 2: Max Planck Institute of Psychiatry, München, Germany

Advancing Sea Ice Surface Classification by Self-Supervised Contrastive Learning for Radar Altimetry

Lena Happ^1,5, Stefan Hendricks¹, Conrad M Albrecht^2,4, Lars Kaleschke¹, Sonali Patil³, Riccardo Fellegara³, Dirk A. Lorenz⁵

1: Section for Sea Ice Physics, Alfred Wegener Institute, Germany; 2: Earth Observation Center, German Aerospace Center, Germany; 3: Institute of Software Technology, German Aerospace Center, Germany; 4: Earth and Environmental Engineering, Columbia University, USA; 5: Center for Industrial Mathematics, Bremen University, Germany

Analytic model-informed representations for sample-efficient learning of scattering patterns

Stefan Häusler, Michael Wagener, Stephan Förster, Marina Ganeva

Forschungszentrum Jülich GmbH, Germany

Quantum Kernel Alignment for Clinical Data Classification

Florian Heininger¹, Jeanette Miriam Lorenz³, Matthias Weidemüller², Niels Halama¹

1: German Cancer Research Center, Germany; 2: Heidelberg University; 3: Fraunhofer IKS, LMU Munich

Syxplain: An Agentic Framework for Explainable Scientific Hypothesis Generation from Data

Roland Hollós¹, Patrik Süli², Gergő Szabó¹, György Eigner²

1: HUN-REN, Hungary; 2: Obuda University, Hungary

Jülich Neutron Agent (JüNA): Vitess AI Agent

Ahmad Zainul Ihsan, Nicolo Violini, Klaus Lieutenant, Marina Ganeva

Jülich Centre for Neutron Science (JCNS), FZ Jülich, Germany

NEOPIC: HPC Strategies for FNO-DSE in Kinetic Plasma Simulation

Chelsea Maria John¹, Sriramkrishnan Muralikrishnan¹, Jeffrey Kelling², Paul Gibbon¹, Andreas Herten¹

1: Forschungszentrum Jülich, Germany; 2: Helmholtz Zentrum Dresden-Rossendorf, Germany

Diagnosing and Fixing Training Bottlenecks: A Case Study of a Foundation Model

Javad Kasravi¹, Christian Lessig², Timothy Hunter², Stefan Kesselheim¹

1: Jülich Supercomputing Centre, Forschungszentrum Jülich, 52428 Jülich, Germany; 2: ECMWF, 53175 Bonn, Germany

Advancing Explainability as a Differentiator in Machine Learning prediction models for Alzheimer’s Disease

Charles Kaumbutha^1,2, Shubhi Ambast^1,2, Karola Mai^1,2, Karoline Mauer^2,3, Matthias Becker^1,2

1: Modular HPC and AI, German Center for Neurodegenerative Diseases (DZNE), Bonn, Germany; 2: Systems Medicine, German Center for Neurodegenerative Diseases (DZNE) e.V., 53127 Bonn, Germany; 3: PRECISE Platform for Genomics and Epigenomics at DZNE and University of Bonn, 53127 Bonn, Germany

Improving Pedestrian Detection through Temporal Movement Estimations

Yannick Kees, Elena Hoemann, Frank Köster, Sven Hallerbach

German Aerospace Center (DLR), Germany

Neural Networks as Varieties

Ismail Khalfaoui-Hassani, Stefan Kesselheim

Forschungszentrum Jülich, Germany

Surrogate Models for Severe Nuclear Accidents

Paul Kieckhefen^1,2, Jonas Dressner², Charlotte Debus², Markus Götz^1,2

1: Helmholtz AI, Germany; 2: Karlsruhe Institute of Technology, Germany

Training-free Uncertainty Guidance for Complex Visual Tasks with MLLMs

Sanghwan Kim^1,2,3, Rui Xiao^2,3, Stephan Alaniz⁴, Yongqin Xian⁵, Zeynep Akata^1,2,3

1: Helmholtz Zentrum München, Germany; 2: Technical University of Munich, Germany; 3: Munich Center for Machine Learning, Germany; 4: Télécom Paris, France; 5: Google, Switzerland

Helixer – a tool for ab initio gene calling combining Deep Learning and a Hidden Markov Model

Felicitas Kindel^1,2, Anthony Bolger¹, Björn Usadel^1,2,3, Alisandra Denton^2,3,4, Marie Bolger¹

1: Forschungszentrum Jülich, Germany; 2: Heinrich Heine Univsersität Düsseldorf, Germany; 3: Cluster of Excellence on Plant Sciences (CEPLAS), Germany; 4: Recursion, Valence Labs, Canada

metabeta - A fast neural model for Bayesian mixed-effects regression

Alex Kipnis, Marcel Binz, Eric Schulz

Human-Centered AI, Helmholtz Munich, Germany

Meta-Learning for Phenotypic Drug Discovery with Transformer Neural Processes

Michal Kmicikiewicz^1,2, Tommy Rochussen^2,3,4, Vincent Fortuin^3,4,6, Ewa Szczurek^1,5

1: Institute of AI for Health, Helmholtz Munich; 2: School of Computation, Information and Technology, Technical University of Munich; 3: Helmholtz AI; 4: Munich Center for Machine Learning; 5: Faculty of Mathematics, Informatics and Mechanics, University of Warsaw; 6: Department of Computer Science and Artificial Intelligence, University of Technology Nuremberg

Agentic Data Extraction and Dynamic Knowledge Graph Construction for Hydrogen Technology Research

Marjan Kohandani^1,2,3, Mohammad j. Eslamibidgoli^1,2, Michael H. Eikerling^1,2,3, Kourosh Malek^1,2

1: Theory and Computation of Energy Materials (IET-3), Institute of Energy Technologies, Forschungszentrum Jülich GmbH, 52425 Jülich, Germany; 2: Centre for Advanced Simulation and Analytics (CASA), Simulation and Data Science Lab for Energy Materials (SDL-EM), Forschungszentrum Jülich GmbH, 52425 Jülich, Germany; 3: Chair for Theory and Computation of Energy Materials, Faculty of Georesources and Materials Engineering, RWTH Aachen University, Germany

Diagnosing Shortcuts in Traffic Video Question Answering Datasets

Sena Korkut¹, María Alejandra Bravo^1,2,3, Sanghwan Kim^1,2,3, Zeynep Akata^1,2,3

1: Technical University of Munich, Germany; 2: Helmholtz Zentrum München, Germany; 3: Munich Center for Machine Learning, Germany

DREAMS: Preserving both Local and Global Structure in Dimensionality Reduction

Noël Kury^1,2, Dmitry Kobak^1,2, Sebastian Damrich^1,2

1: Hertie Institute for AI in Brain Health, Germany; 2: University of Tübingen, Germany

Factor dependencies and their associations with covariates in multimodal factor analysis

Małgorzata Łazęcka¹, Ewa Szczurek^2,1

1: University of Warsaw, Poland; 2: Helmholtz Center Munich, Germany

Discovering Interpretable Visual Concepts in Monkey Visual Cortex Using Sparse Autoencoders

Lilly May^1,2, Atharv Suryawanshi¹, Steffen Schneider¹

1: Institute of Computational Biology, Computational Health Center, Helmholtz Munich, Germany; 2: Department of Experimental Psychology, University of Oxford, UK

Incremental learning for continuous forest disturbance mapping using AlphaEarth embeddings

Francisco Mena, Simon Besnard, Martin Herold

GFZ Helmholtz Centre for Geosciences, Germany

Multimodal Machine Learning for Postoperative Atrial Fibrillation Prediction

Julian Mierisch¹, Christian Niklas², Rosa Klotz³, Axel Loewe¹

1: Karlsruhe Institute of Technology, Germany; 2: Medical Faculty Mannheim, Germany; 3: University Hospital Heidelberg, Germany

FrevaGPT: An LLM-Powered Assistant for Reproducible Climate Analysis

Koketso Michelle Molepo¹, Gizem Ekinci², Sebastian Willmann², Johanna Baehr³, Kevin Sieck⁴, Felix Oertel², Bianca Wentzel², Thomas Ludwig², Martin Bergemann², Jan Saynisch-Wagner¹, Christopher Kadow²

1: GFZ Helmholtz Centre for Geosciences, Germany; 2: German Climate Computing Center DKRZ, Germany; 3: Universität Hamburg, Germany; 4: Climate Service Center Germany (GERICS), Germany

Mind The Gap: Continuous Magnification Sampling For Pathology Foundation Models

Alexander Möllers^1,2,3, Julius Hense^1,2, Florian Schulz^1,2, Timo Milbich³, Maximilian Alber³, Lukas Ruff^3,4

1: Technical University Berlin, Germany; 2: Berlin Institute for the Foundations of Learning and Data; 3: Aignostics Gmbh; 4: Charité, Universitätsmedizin Berlin

Grounding Large Language Models in Formal Solvers via Unified Planning: Towards High-Accuracy Natural Language Planning

João Areias Saraiva, Thomas Kirste

Hybrid Methods in Artificial Intelligence and Machine Learning, University of Rostock, Germany

MLArray: Unifying Machine Learning-Optimized Array Storage and Imaging Metadata

Karol Gotkowski, Kris Dreher

German Cancer Research Center (DKFZ), Germany

Post-training makes large language models less human-like

Marcel Binz, Psych-201 Consortium, Eric Schulz

Helmholtz Munich, Germany

A Domain-Specialized Multimodal Foundation Model for Interpreting Sentinel-2 True Color Imagery in Flood Disaster Response

Luca Chiarabini, Daniela Espinoza Molina, Andrés Camero Unzueta

DLR, Germany

A Hematology Foundation Model Enables Blood Cancer Screening from Peripheral Blood Smears

Muhammed Furkan Dasdelen¹, Fatih Ozlugedik¹, Ivan Kukuljan¹, Ario Sadafi¹, Matthias Hehr¹, Christian Pohlkamp², Carsten Marr¹

1: Helmholtz Munich, Germany; 2: Munich Leukemia Laboratory

Surrogate Modeling of Scale-Bridging Dislocation-Driven Plasticity through Latent Dynamics and Microstructure Descriptors

Aytekin Demirci¹, Stefan Sandfeld^1,2

1: Institute for Advanced Simulations – Materials Data Science and Informatics (IAS-9) Forschungszentrum Jülich GmbH 52425 Jülich, Germany; 2: Chair of Materials Data Science and Materials Informatics, Faculty 5 Georesources and Materials Engineering, RWTH Aachen University 52056 Aachen, Germany

NVIDIA Earth-2: Open Models, Accessible Hardware, and the Future of AI for Weather and Climate

David Hall, Mike Pritchard, Karthik Kashinath, Noah Brenowitz

NVIDIA Corporation

Amortized Bayesian Multilevel Model of Bacterial Cell-Cycle with BayesFlow

Laura Jenniches¹, Lars Barquist^1,2,3

1: Helmholtz Institute for RNA-based Infection Research (HIRI), Helmholtz Centre for Infection Research (HZI), Würzburg, Germany; 2: Department of Biology, University of Toronto, Mississauga, ON, Canada; 3: Department of Cell and Systems Biology, University of Toronto, Toronto, ON, Canada

Exploring the Effectiveness of Pretraining Task Alignment for Self-Supervised Learning in a Low-Data Regime

Shivangi Kirti¹, Binh Duong Nguyen¹, Stefan Sandfeld^1,2

1: Institute for Advanced Simulations — Materials Data Science and Informatics (IAS-9), Forschungszentrum Jülich GmbH, 52425 Jülich, Germany; 2: Chair of Materials Data Science and Materials Informatics, Faculty 5 — Georesources and Materials Engineering, RWTH Aachen University, 52056 Aachen, Germany

Taming Multimodal Posteriors over Rotations: A Geometric Approach to Bayesian Alignment in Cryo-EM and Beyond

Shantanu Kodgirwar^1,2,3, Jan Rothhardt^1,2,3,4, Michael Habeck^3,5,6

1: Helmholtz Institute Jena, Germany; 2: GSI Helmholtzzentrum für Schwerionenforschung, Germany; 3: Friedrich Schiller University Jena, Germany; 4: Fraunhofer Institute for Applied Optics and Precision Engineering, Germany; 5: University Hospital Jena, Germany; 6: Max Plank Institute for Multidisciplinary Sciences, Germany

Sample Efficient Generative Molecular Optimization with Joint Self-Improvement

Serra Korkmaz¹, Adam Izdebski^1,2, Jonathan Pirnay^3,4, Rasmus Møller-Larsen^1,2, Michal Kmicikiewicz^1,2, Pankhil Gawade^1,2, Dominik G. Grimm^2,3,4, Ewa Szczurek^1,5

1: Helmholtz Munich, Germany; 2: Technical University of Munich, TUM School of Computation, Information and Technology; 3: Technical University of Munich, Campus Straubing for Biotechnology and Sustainability; 4: University of Applied Sciences Weihenstephan Triesdorf; 5: Faculty of Mathematics, Informatics and Mechanics, University of Warsaw

Mind the Gap – From expectation to reality in agentic AI

Peyman Vahidi^1,2, Sebastian Lobentanzer^1,2

1: Helmholtz Munich, Germany; 2: German Center for Diabetes Research, Germany

Genomic language model based prediction of transcription start sites in twelve yeast species

Maria Ryabtseva¹, Tilman Hoffbauer¹, Julien Gagneur^1,2

1: Computation Molecular Medicine, School of Computation, Information and Technology, TU Munich, Germany; 2: Computational Health Center, Helmholtz Center Munich, Germany