Samuel Denton

Sebastian Salazar, Samuel Denton, Ansaf Salleb-Aouissi

We tackle the problem of computing counterfactual explanations -- minimal changes to the features that flip an undesirable model prediction. We propose a solution to this question for linear Support Vector Machine (SVMs) models. Moreover, we introduce a way to account for weighted actions that allow for more changes in certain features than others. In particular, we show how to find counterfactual explanations with the purpose of increasing model interpretability. These explanations are valid, change only actionable features, are close to the data distribution, sparse, and take into account correlations between features. We cast this as a mixed integer programming optimization problem. Additionally, we introduce two novel scale-invariant cost functions for assessing the quality of counterfactual explanations and use them to evaluate the quality of our approach with a real medical dataset. Finally, we build a support vector machine model to predict whether law students will pass the Bar exam using protected features, and used our algorithms to uncover the inherent biases of the SVM.

Julia Shuieh, Prasann Singhal, Apaar Shanker et al.

Supervised and preference-based fine-tuning techniques have become popular for aligning large language models (LLMs) with user intent and correctness criteria. However, real-world training data often exhibits spurious correlations -- arising from biases, dataset artifacts, or other "shortcut" features -- that can compromise a model's performance or generalization. In this paper, we systematically evaluate three post-training algorithms -- Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and KTO (Kahneman-Tversky Optimization) -- across a diverse set of synthetic tasks and spuriousness conditions. Our tasks span mathematical reasoning, constrained instruction-following, and document-grounded question answering. We vary the degree of spurious correlation (10% vs. 90%) and investigate two forms of artifacts: "Feature Ambiguity" and "Distributional Narrowness." Our results show that the models often but not always degrade under higher spuriousness. The preference-based methods (DPO/KTO) can demonstrate relative robustness in mathematical reasoning tasks. By contrast, SFT maintains stronger performance in complex, context-intensive tasks. These findings highlight that no single post-training strategy universally outperforms in all scenarios; the best choice depends on the type of target task and the nature of spurious correlations.