<ul><li>Conformal Prediction (CP) is effective for improving the trustworthiness of neural networks by providing prediction sets with finite-sample guarantees.</li><li>Robust Conformal Prediction addresses the issue of classical conformal guarantees not holding under adversarial attacks.</li><li>A new method leveraging Lipschitz-bounded networks has been proposed to estimate robust CP sets efficiently and precisely.</li><li>This method, called lip-rcp, outperforms existing approaches in terms of robust CP set size and computational efficiency in large-scale scenarios like ImageNet.</li></ul>

Efficient Robust Conformal Prediction via Lipschitz-Bounded Networks

Discover more