Acknowledgements

Julyan Arbel acknowledges support from ANR-21-JSTM-0001 and ANR-23-IACL-0006 grants.

Battista Biggio acknowledge support from the EU-funded Horizon Europe project ELSA (GA no. 101070617) the projects SERICS (PE00000014) and FAIR (PE00000013) under the MUR National Recovery and Resilience Plan funded by the European Union - NextGenerationEU. Fabio Roli acknowledge support from the project FISA-2023-00128 funded by the MUR program “Fondo italiano per le scienze applicate”.

Michele Caprio acknowledges partial funding by the Army Research Office (ARO MURI W911NF2010080).

The authors acknowledge financial support from Imperial College London through an Imperial College Research Fellowship grant awarded to Gergely Flamich.

Vincent Fortuin was supported by a Branco Weiss Fellowship.

Philipp Hennig, Nathaël da Costa, Tobias Weber, Bálint Mucsányi, and Joanna Sliwa gratefully acknowledge financial support by the European Research Council through ERC CoG Action 101123955 ANUBIS; the DFG Cluster of Excellence “Machine Learning - New Perspectives for Science”, EXC 2064/1; the DFG SPP 2298 (Project HE 7114/5-1), and the Carl Zeiss Foundation, (project “Certification and Foundations of Safe Machine Learning Systems in Healthcare”), as well as funds from the Ministry of Science, Research and Arts of the State of Baden-Württemberg.

Christian Igel acknowledges support from the Danish National Research Foundation (DNRF) through the Danish Pioneer Centre for AI (P1) and TreeSense, the Center for Remote Sensing and Deep Learning of Global Tree Resources (DNRF192).

Nadja Klein (N. K.) and Clara Hoffmann (C. H.) would like to thank Moussa Kassem Sbeyti for his helpful feedback. C. H. would like to thank Katja Weimer for her valuable comments and suggestions. N. K. and C. H. acknowledge funding by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) through the grant KL3037/8-1 and the Emmy Noether grant KL 3037/1-1.

Jakob Macke and Michael Deistler were supported by the German Research Foundation (DFG) under Germany’s Excellence Strategy (EXC number 2064/1; 390727645) and SFB 1233 `Robust Vision’ (276693517), the European Union (ERC DeepCoMechTome; 101089288), and the Hector Foundation.

Dhruvesh Patel acknowledges support from IBM under IBM Research Collaboration Agreement No. W1668553 and from the National Science Foundation under grant IIS-2106391.

Tom Rainforth was supported by the EPSRC grant EP/Y037200/1.

Freddie Bickford Smith was supported by the EPSRC Probabilistic AI Hub (EP/Y028783/1).