KAN-Robust-Bench: A Benchmark for Evaluating the Robustness of Kolmogorov-Arnold Networks
Researchers have proposed a new benchmark for evaluating the robustness of Kolmogorov-Arnold Networks (KANs) against adversarial attacks. The KAN-Robust-Bench assesses the certified and empirical robustness of various KAN architectures under strong evasion attacks, including FGSM, PGD, and C&W attacks. The authors provide mathematical foundations for randomized smoothing and interval bound propagation to evaluate the models' robustness.
Researchers have proposed a new benchmark for evaluating the robustness of Kolmogorov-Arnold Networks (KANs) against adversarial attacks. The KAN-Robust-Bench assesses the certified and empirical robustness of various KAN architectures under strong evasion attacks, including FGSM, PGD, and C&W attacks. The authors provide mathematical foundations for randomized smoothing and interval bound propagation to evaluate the models' robustness.
---
Why it matters: This matters because many machine learning models are vulnerable to adversarial attacks, which can compromise their security. Evaluating the robustness of KANs is crucial for developing secure AI systems that can withstand such threats.
Source: https://arxiv.org/abs/2608.21488
This article was originally published at: https://arxiv.org/abs/2608.21488