aiqbench analysis
SAGE addresses the critical issue of spurious correlations in machine learning by using generative models to synthesize data for underrepresented subpopulations, thereby improving worst-group accuracy without requiring prior knowledge of group labels. This approach offers a robust solution for building more equitable AI systems that generalize better across diverse datasets like Waterbirds and CelebA.
SAGE: Subpopulation-Aware Generative Enhancement for Mitigating Spurious Correlations
arXiv:2609.01051v1 Announce Type: new Abstract: Spurious correlations pose a significant challenge to the robustness of modern machine learning. The inherent imbalance in dataset distributions often leads traditional Empirical Risk Minimization (ERM) models to rely on majority spurious attributes for classification, resulting in poor performance on minority groups. This problem becomes particularly challenging when the spurious attributes are unavailable. Existing group-label-free methods often upsample minority groups or misclassified real training examples; repeating the same instances can reduce effective diversity and encourage overfitting. To mitigate these spurious correlations from a data-centric perspective in the absence of prior knowledge, we introduce Subpopulation-Aware Generative Enhancement (SAGE), a two-stage generative augmentation framework. Using cluster-derived sub-labels and class labels, we fine-tune a conditional generative model and text encoder, generating targeted synthetic data to fill underrepresented regions in the training set and construct a balanced validation set for last-layer reweighting. We experimentally show that SAGE achieves 89.5%, 85.7%, and 79.1% worst-group accuracy on Waterbirds, CelebA, and MetaShift, respectively, outperforming the best group-label-free baselines by up to 7.7 percentage points.
Abstract:Spurious correlations pose a significant challenge to the robustness of modern machine learning. The inherent imbalance in dataset distributions often leads traditional Empirical Risk Minimization (ERM) models to rely on majority spurious attributes for classification, resulting in poor performance on minority groups. This problem becomes particularly challenging when the spurious attributes are unavailable. Existing group-label-free methods often upsample minority groups or misclassified real training examples; repeating the same instances can reduce effective diversity and encourage overfitting. To mitigate these spurious correlations from a data-centric perspective in the absence of prior knowledge, we introduce Subpopulation-Aware Generative Enhancement (SAGE), a two-stage generative augmentation framework. Using cluster-derived sub-labels and class labels, we fine-tune a conditional generative model and text encoder, generating targeted synthetic data to fill underrepresented regions in the training set and construct a balanced validation set for last-layer reweighting. We experimentally show that SAGE achieves 89.5%, 85.7%, and 79.1% worst-group accuracy on Waterbirds, CelebA, and MetaShift, respectively, outperforming the best group-label-free baselines by up to 7.7 percentage points.
Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
arXiv:2609.01051 [cs.LG]
(or arXiv:2609.01051v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.01051 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 267 words
Related coverage
Article: Artificial Intelligence-Driven Phishing: How Phishing Technique Is Evolving and Implemented
InfoQ · Jun 8, 2026
OnlineWM: Causality-Aware Active Online Learning for Effective World Modeling
arXiv · Sep 22, 2026
Why do models task game?
AI Alignment Forum · Aug 6, 2026
Cross-reality location privacy protection in 6G-enabled vehicular metaverses: an LLM-enhanced hybrid generative diffusion model-based approach
arXiv · Jul 27, 2026