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.

We revisit survey results from Section 6, and explore the impact of label noise on our proposed method. Each participant was told that an auction would be fair if each ad was presented to each group at equal rates. We train PreferenceNet, perturbing labels according to their distance from the decision boundary. Eliciting preferences from a group is particularly challenging, because these preferences often heterogeneous. Table 2: We simulate preference elicitation of three different definitions of fairness.