UGCA increases Expected Calibration Error of GNNs under adversarial edge perturbations while preserving classification accuracy, with theoretical links between model accuracy, dataset complexity, and vulnerability.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
The Confidence Trap: Calibration Attacks for Graph Neural Networks
UGCA increases Expected Calibration Error of GNNs under adversarial edge perturbations while preserving classification accuracy, with theoretical links between model accuracy, dataset complexity, and vulnerability.