Vaclav Voracek

Researcher, SF

Joined

September 2021

Names

Vaclav Voracek (Preferred)

Emails

****@uni-tuebingen.de (Confirmed)

,

****@gmail.com (Confirmed)

Personal Links

Career & Education History

Researcher

SF (second-foundation.eu)

2025 – Present

Advisors, Relations & Conflicts

Coauthor

2024 – Present

Coauthor

Moise Blanchard

2023 – Present

PhD Advisor

2021 – Present

Expertise

Certified Adversarial Robustness

2021 – Present

Publications

STAR-Bets: Sequential TArget-Recalculating Bets for Tighter Confidence Intervals
Vaclav Voracek, Francesco Orabona
- NeurIPS 2025 poster
- Readers: Everyone
An Interpretable N-gram Perplexity Threat Model for Large Language Model Jailbreaks
Valentyn Boreiko, Alexander Panfilov, Vaclav Voracek, Matthias Hein, Jonas Geiping
- ICML 2025 poster
- Readers: Everyone
A Realistic Threat Model for Large Language Model Jailbreaks
Valentyn Boreiko, Alexander Panfilov, Vaclav Voracek, Matthias Hein, Jonas Geiping
- Submitted to ICLR 2025
- Readers: Everyone
A Realistic Threat Model for Large Language Model Jailbreaks
Valentyn Boreiko, Alexander Panfilov, Vaclav Voracek, Matthias Hein, Jonas Geiping
- Red Teaming GenAI Workshop @ NeurIPS'24 Oral
- Readers: Everyone
Treatment of Statistical Estimation Problems in Randomized Smoothing for Adversarial Robustness
Vaclav Voracek
- NeurIPS 2024 poster
- Readers: Everyone
Convergence of Some Convex Message Passing Algorithms to a Fixed Point
Vaclav Voracek, Tomas Werner
- ICML 2024 Spotlight
- Readers: Everyone
Improving l1-Certified Robustness via Randomized Smoothing by Leveraging Box Constraints
Vaclav Voracek, Matthias Hein
- ICML 2023 Poster
- Readers: Everyone
Sound Randomized Smoothing in Floating-Point Arithmetic
Vaclav Voracek, Matthias Hein
- Published: 01 Feb 2023, Last Modified: 14 Jan 2026
- ICLR 2023 poster
- Readers: Everyone
Provably Adversarially Robust Nearest Prototype Classifiers
Václav Vorácek, Matthias Hein
- 2022 (modified: 17 Jan 2023)
- ICML 2022
- Readers: Everyone

Co-Authors