Leon Mayer

I’m a PhD student at the German Cancer Research Center (DKFZ) working on benchmarking and understanding vision-language models in the Group for Intelligent Medical Systems. I am currently most interested in building meaningful evals and in agent monitorability.

I work on this because I expect AI to become increasingly consequential, and I want to help make sure it goes well for all sentient beings.

Before my PhD, I earned a BSc in Electrical Engineering from RWTH Aachen and an MSc from TU Munich. I’ve also volunteered with ARMoR, a nonprofit advocating for policies to fight antimicrobial resistance and, in summer 2026, completed BlueDot Impact’s AGI Strategy Course.

If you find any of the above interesting, feel free to reach out!

news

Jun 02, 2026	We presented our preprint 6 Fingers, 1 Kidney: Natural Adversarial Medical Images Reveal Critical Weaknesses of Vision-Language Models (arXiv) as an Oral during the MMFM-BIOMED workshop at CVPR 2026.
Feb 04, 2026	Our surgical foundation model SurgVISTA has been published in npj Digital Medicine.
Jan 20, 2026	I gave an invited talk about the challenges of building benchmarks for AI in medicine at the journal club of the Optimization of Machine Learning Systems Group at the University of Basel.

selected publications

arXiv

HeiCo-FOCUS: A Clinically Grounded Dataset for Long-Context Video Understanding

Leon Mayer, Lucas Luttner, Patrick Godau, and 42 more authors

arXiv preprint, 2026

arXiv Bib Code

@article{mayer2026heico,
  title = {HeiCo-FOCUS: A Clinically Grounded Dataset for Long-Context Video Understanding},
  author = {Mayer, Leon and Luttner, Lucas and Godau, Patrick and Fritzsche, Kai and Reinke, Annika and Boland, Leonie and Brandt, Jule and Heinecke, Janne and Nobuhara, Chloe K and Holzwarth, Niklas and Christodoulou, Evangelia and Knopp, Marcel and Michael, Dominik and Piermarco, Pascale and Neyaz, Saliq and {\"O}zarslan, Korhan Derin and Hennighausen, Jakob and Aumente-Maestro, Carlos and R{\"a}dsch, Tim and Baji, Dheeraj and Full, Peter Maximilian and Aichholz, Finn and Erpenbeck, Justus Veit and Schott, Linus Finn and Winkelhausen, Bastian and de Boer, Claas and G{\"u}ttner, Bianca and Hummel, Anneli and Just, Gregor and Kirchner, Max and Li, Chenyang and Raffaut, Rozenn and Rodriguez, Ariel and Venkatesh, Danush Kumar and Wang, Kevin and Xu, Jinjing and Sheikh Zeinoddin, Mona and Khan, Salman and Pausch, Thomas M and Speidel, Stefanie and Stoyanov, Danail and Hashimoto, Daniel A and Kolbinger, Fiona R and Weiser, Thomas G and Maier-Hein, Lena},
  journal = {arXiv preprint},
  year = {2026},
}

npj Digit Med

Large-scale Self-supervised Video Foundation Model for Intelligent Surgery

Shu Yang, Fengtao Zhou, Leon Mayer, and 16 more authors

npj Digital Medicine, 2026

arXiv Bib HTML Code

@article{yang2025large,
  title = {Large-scale Self-supervised Video Foundation Model for Intelligent Surgery},
  author = {Yang, Shu and Zhou, Fengtao and Mayer, Leon and Huang, Fuxiang and Chen, Yiliang and Wang, Yihui and He, Sunan and Nie, Yuxiang and Wang, Xi and S{\"u}mer, {\"O}mer and Jin, Yueming and Sun, Huihui and Xu, Shuchang and Liu, Alex Qinyang and Li, Zheng and Qin, Jing and Teoh, Jeremy YuenChun and Maier-Hein, Lena and Chen, Hao},
  journal = {npj Digital Medicine},
  volume = {9},
  number = {1},
  pages = {220},
  year = {2026},
  publisher = {Nature Publishing Group UK London},
}

arXiv

6 Fingers, 1 Kidney: Natural Adversarial Medical Images Reveal Critical Weaknesses of Vision-Language Models

Leon Mayer, Piotr Kalinowski, Caroline Ebersbach, and 6 more authors

arXiv preprint, 2025

arXiv Bib Code

@article{mayer20256,
  title = {6 Fingers, 1 Kidney: Natural Adversarial Medical Images Reveal Critical Weaknesses of Vision-Language Models},
  author = {Mayer, Leon and Kalinowski, Piotr and Ebersbach, Caroline and Knopp, Marcel and R{\"a}dsch, Tim and Christodoulou, Evangelia and Reinke, Annika and Kolbinger, Fiona R and Maier-Hein, Lena},
  journal = {arXiv preprint},
  year = {2025},
}

arXiv

Challenging Vision-Language Models with Surgical Data: A New Dataset and Broad Benchmarking Study

Leon Mayer, Tim Rädsch, Dominik Michael, and 8 more authors

arXiv preprint, 2025

arXiv Bib

@article{mayer2025challenging,
  title = {Challenging Vision-Language Models with Surgical Data: A New Dataset and Broad Benchmarking Study},
  author = {Mayer, Leon and R{\"a}dsch, Tim and Michael, Dominik and Luttner, Lucas and Yamlahi, Amine and Christodoulou, Evangelia and Godau, Patrick and Knopp, Marcel and Reinke, Annika and Kolbinger, Fiona and Maier-Hein, Lena},
  journal = {arXiv preprint},
  year = {2025},
}

Comms Med

Robotic scrub nurse to anticipate surgical instruments based on real-time laparoscopic video analysis

Lars Wagner, Sara Jourdan, Leon Mayer, and 10 more authors

Communications Medicine, 2024

Bib HTML PDF

@article{wagner2024robotic,
  title = {Robotic scrub nurse to anticipate surgical instruments based on real-time laparoscopic video analysis},
  author = {Wagner, Lars and Jourdan, Sara and Mayer, Leon and M{\"u}ller, Carolin and Bernhard, Lukas and Kolb, Sven and Harb, Farid and Jell, Alissa and Berlet, Maximilian and Feussner, Hubertus and Buxmann, Peter and Knoll, Alois and Wilhelm, Dirk},
  journal = {Communications Medicine},
  volume = {4},
  number = {1},
  pages = {156},
  year = {2024},
  publisher = {Nature Publishing Group UK London},
}

Sensors

A deep learning-based camera approach for vital sign monitoring using thermography images for ICU patients

Simon Lyra, Leon Mayer, Liyang Ou, and 7 more authors

Sensors, 2021

Bib HTML PDF

@article{lyra2021deep,
  title = {A deep learning-based camera approach for vital sign monitoring using thermography images for ICU patients},
  author = {Lyra, Simon and Mayer, Leon and Ou, Liyang and Chen, David and Timms, Paddy and Tay, Andrew and Chan, Peter Y and Ganse, Bergita and Leonhardt, Steffen and Hoog Antink, Christoph},
  journal = {Sensors},
  volume = {21},
  number = {4},
  pages = {1495},
  year = {2021},
  publisher = {MDPI},
}