Twelve papers co-authored by BIFOLD researchers span core areas of machine learning at ICML 2026
The Forty-Third International Conference on Machine Learning (ICML 2026) takes place from July 6 to 11, 2026, in Seoul, Korea. ICML is a premier gathering dedicated to advancing machine learning, presenting and publishing cutting-edge research across artificial intelligence, statistics, and data science, as well as application areas such as machine vision, computational biology, speech recognition, and robotics. Participants range from academic and industrial researchers to entrepreneurs, engineers, graduate students, and postdocs.
This year, twelve papers are co-authored by BIFOLD researchers and/or were developed in close collaboration with BIFOLD researchers together with researchers from partner institutions like Fraunhofer Heinrich Hertz Institute (HHI) and Riken Institute. Six research papers will be presented at the main conference, and six are workshop papers. Among these, “Learning Hamiltonian Flow Maps: Mean Flow Consistency for Large-Timestep Molecular Dynamics” received a Spotlight award, placing it among the top 2.2 percent of all submissions.
Below is an overview of BIFOLD’s contributions:
Beyond the Final Layer: Attentive Multilayer Fusion for Vision Transformers. Laure Ciernik, Marco Morik, Lukas Thede, Luca Eyring, Shinichi Nakajima, Zeynep Akata, Lukas Muttenthaler.
- Paper
- Wed, Jul 8, Poster Session 3, 10:30 AM to 12:15 PM
AlgoTrace: Algorithmic Primitives and Compositional Geometry of Reasoning in Language Models.Samuel Lippl, Thomas Austin McGee, Kimberly Lopez, Ziwen Pan, Pierce Zhang, Salma Ziadi, Oliver Eberle, Ida Momennejad.
- Paper
- Hall A. Wed, Jul 8, 2026 • 10:30 PM – 12:15 AM
EEG-Based Multimodal Learning via Hyperbolic Mixture-of-Curvature Experts. Runhe Zhou, Shanglin Li, Guanxiang Huang, Xinliang Zhou, Qibin Zhao, Motoaki Kawanabe, Yi Ding, Cuntai Guan.
Incorporating Importance Weighting in Optimal Transport Based Domain Alignment. Okan Koç, Alexander Soen, Shanglin Li, Masashi Sugiyama.
- Hall A #4614. Wed, Jul 8, 2026
Learning Hamiltonian Flow Maps: Mean Flow Consistency for Large-Timestep Molecular Dynamics. Winfried Ripken, Michael Plainer, Gregor Lied, Thorben Frank, Oliver T. Unke, Stefan Chmiela, Frank Noé, Klaus Robert Müller.
- Paper | Website | Code
- Hall A. Wed, Jul 8, 2026 • 10:30 AM – 12:15 PM
- Awarded a Spotlight, representing the top 2.2% of all submissions.
PINNfluence: Interpreting PINNs through Influence Functions. Aleksander Krasowski, Jonas R. Naujoks, Moritz Weckbecker, Galip Ümit Yolcu, Thomas Wiegand, Sebastian Lapuschkin, Wojciech Samek, René P. Klausen.
- Paper | Website (demo) | Code
- Hall A. Wed, Jul 8, 2026 • 10:30 AM – 12:15 PM
Federated Sampling of Molecular Conformers via Compositional Flows. Dennis Grinwald, Philipp Wiesner, Shinichi Nakajima.
- July 10, ICML 2026 Workshop on Structured Probabilistic Inference & Generative Modeling (SPIGM)
- Paper
Activation- and Influence-Aware Ranks (AIR): Function-Preserving SVD Compression for LLMs. Nico Harder, Daniel Becking, Karsten Müller, Wojciech Samek.
- July 10, ICML Workshop on “Resource-Adaptive Foundation Model Inference (AdaptFM),” 2026.
LieSolver: A PDE-Constrained Solver for IBVPs Using Lie Symmetries. René P. Klausen, Ivan Timofeev, Johannes Frank, Jonas Naujoks, Thomas Wiegand, Sebastian Lapuschkin, Wojciech Samek.
- July 11, ICML Workshop on “AI for Physics,” 2026.
- Paper
Predicting Future Behaviors in Reasoning Models Enables Better Steering. Evgenii Kortukov, Piotr Komorowski, Florian Klein, Paula Engl, Gabriele Sarti, Seong Joon Oh, Sebastian Lapuschkin, Wojciech Samek.
- July 11, ICML Workshop on “Mechanistic Interpretability,” 2026.
- Paper
From Reward-Hack Activations to Agentic Risk States: Context-Calibrated Mechanistic Monitoring in LLM Agents. Patrick Wilhelm, Odej Kao.
- July 11, ICML Workshop on “Agents in the Wild: Safety, Security, and Beyond.”
- Paper
Thought Virus: Viral Misalignment via Subliminal Prompting in Multi-Agent Systems. Moritz Weckbecker, Jonas Müller, Ben Hagag, Michael Mulet.