Scholar
Francesco Croce
Google Scholar ID: laq9cq0AAAAJ
EPFL
Citations & Impact
Citations
7,680
H-index
23
i10-index
27
Publications
20
Co-authors
16
list available
Publications
Resume (English only)
Academic Achievements
- DAGM MVTec Dissertation Award 2024; Wilhelm Schickard Dissertation Award 2024; 1st place at the Find the Trojan: Universal Backdoor Detection in Aligned LLMs competition; 5th most influential paper of ICML 2020 'Reliable evaluation of adversarial robustness with an ensemble of diverse and parameter-free attacks'; Best Paper Honorable Mention Award for 'RobustBench: a standardized adversarial robustness benchmark' at ICLR 2021 Workshop on Security and Safety in ML Systems; Honorable Mention Award for 'A randomized gradient-free attack on ReLU networks' at GCPR 2018; Multiple papers accepted or published in NeurIPS 2025, ICLR 2025, SaTML 2025.
Research Experience
- During his PhD, he mainly worked on adversarial robustness in vision tasks, developing AutoAttack and RobustBench. His current research focuses on the adversarial robustness of multimodal foundation models, especially for safe AI systems. He is also exploring their ability to capture different aspects of human perception, such as visual and semantic similarity.
Education
- PhD in Computer Science: 2023, University of Tübingen, supervised by Prof. Matthias Hein; BSc and MSc in Mathematics: University of Torino; Interned at DeepMind London in 2022, hosted by Sven Gowal.
Background
- Research Interests: Multimodal foundation models, adversarial robustness (jailbreaks, backdoors, etc.), visual reasoning; Professional Field: Computer Science, Machine Learning; Brief Introduction: Currently an Assistant Professor (Tenure-Track) at Aalto University, PS Fellow, and PI at the ELLIS Institute Finland.
Miscellany
- Recruiting students through the ELLIS PhD Program; Interested in multimodal learning, robustness, visual reasoning (and more), feel free to reach out.
