We evaluated adversarial robustness with CAA, an ensemble of gradient and search attacks which was recently demonstrated as the most effective attack against a tabular model.

Thisresult shows that visual grounding/feature attribution can be ineffective at resolving shortcuts.

However, the combinatorial number of variable subsets prevents direct application of typical RL algorithms.

Existing methods learn deterministic mappings without explicitly modelling the robustness to outliers or predictive uncertainty,leading to performance degradation when encountering unseen perturbations attest time. Toaddress this, we propose anovelprobabilistic method based on Uncertainty-aware Generalized AdaptiveCycle Consistency(UGAC), which models the per-pixel residual by generalized Gaussian distribution, capable of modelling heavy-tailed distributions.