Efficient Robust Conformal Prediction via Lipschitz-Bounded Networks
Thomas Massena, Léo Andéol, Thibaut Boissin, Franck Mamalet, Corentin Friedrich, Mathieu Serrurier, Sébastien Gerchinovitz
Abstract
Conformal Prediction (CP) has proven to be an effective post-hoc method for improving the trustworthiness of neural networks by providing prediction sets with finite-sample guarantees. However, under adversarial attacks, classical conformal guarantees do not hold anymore: this problem is addressed in the field of Robust Conformal Prediction. Several methods have been proposed to provide robust CP sets with guarantees under adversarial perturbations, but, for large scale problems, these sets are either too large or the methods are too computationally demanding to be deployed in real life scenarios. In this work, we propose a new method that leverages Lipschitzbounded networks to precisely and efficiently estimate robust CP sets. When combined with a 1-Lipschitz robust network, we demonstrate that our lip-rcp method outperforms state-of-the-art results in both the size of the robust CP sets and computational efficiency in medium and largescale scenarios such as ImageNet. Taking a different angle, we also study vanilla 1 CP under attack, and derive new worst-case coverage bounds of vanilla CP sets, which are valid simultaneously for all adversarial attack levels. Our lip-rcp method makes this second approach as efficient as vanilla CP while also allowing robustness guarantees.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
3 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- Language Models with Conformal Factuality Guarantees cites Class-Conditional Conformal Prediction with Many Classes
- An Adaptive Orthogonal Convolution Scheme for Efficient and Flexible CNN Architectures cites A Dynamical System Perspective for Lipschitz Neural Networks
Cited by top-tier papers3
- One Sample is Enough to Make Conformal Prediction RobustSoroush H. Zargarbashi, Mohammad Sadegh Akhondzadeh, Aleksandar BojchevskiNeurIPS 2025 · 4 citations
- Front-Loaded Robust Conformal Prediction: Heavy Calibration, Minimal Test-Time CostSoroush H. Zargarbashi, Mohammad Sadegh Akhondzadeh, Aleksandar BojchevskiICML 2026
- An Adaptive Orthogonal Convolution Scheme for Efficient and Flexible CNN ArchitecturesThibaut Boissin, Franck Mamalet, Thomas Fel, Agustin Martin Picard et al.ICML 2025
Builds on25
- Towards Evaluating the Robustness of Neural NetworksNicholas Carlini, David A. WagnerS&P 2017 · 9,786 citations
- Class-Conditional Conformal Prediction with Many ClassesTiffany Ding, Anastasios Angelopoulos, Stephen Bates, Michael I. Jordan et al.NeurIPS 2023 · 160 citations
- Globally-Robust Neural NetworksKlas Leino, Zifan Wang, Matt FredriksonICML 2021 · 150 citations
- Language Models with Conformal Factuality GuaranteesChristopher Mohri, Tatsunori HashimotoICML 2024 · 107 citations
- A Dynamical System Perspective for Lipschitz Neural NetworksLaurent Meunier, Blaise Delattre, Alexandre Araujo, Alexandre AllauzenICML 2022 · 69 citations
Related papers
- Verifiably Robust Conformal PredictionLinus Jeary, Tom Kuipers, Mehran Hosseini, Nicola PaolettiNeurIPS 2024 · 16 citations
- Adversarially Robust Conformal PredictionAsaf Gendler, Tsui-Wei Weng, Luca Daniel, Yaniv RomanoICLR 2022 · 51 citations
- Ensemble Conformal Predictor (EnCP): A New Conformal Predictor with Robustness Guarantees Against Data Poisoning AttacksYuxin Yang, Qiang Li, Runyang Feng, Liren Shan et al.S&P 2026
- Provably Robust Conformal Prediction with Improved EfficiencyGe Yan, Yaniv Romano, Tsui-Wei WengICLR 2024 · 26 citations
- Robust Conformal Prediction with a Single Binary CertificateSoroush H. Zargarbashi, Aleksandar BojchevskiICLR 2025
