Modelpedia
Work in progress
About
Findings
Models
Concepts
Methods
Datasets
Sources
Light
Dark
Towards Certification of Uncertainty Calibration under Adversarial Attacks
2025-01-22
· ICLR 2025 Poster ·
anchor
Findings
IC-542
The PyTorch pretrained ResNet50 on ImageNet is vulnerable to (1, y)-ACE calibration attacks that increase ECE from 3.70% to 47.23% while preserving accuracy