When Data Imbalance Helps: Robust Generalization Through Shortcut Saturation
We study robust generalization under spurious correlations: tasks where a shortcut feature is correlated with the true label in training but anti-correlated in an adversarial held-out split. Varying the spurious ratio $r$ (the fraction of training examples where shortcut = true label) and model capacity, we find a counterintuitive result: data imbalance promotes generalization in sufficiently capable models. On a synthetic task where the true label is sum parity of an integer sequence and the shortcut is the parity of the maximum-valued element, a 2-layer, 2-head transformer generalized (reached $100\%$ adversarial accuracy) in 0% of seeds at $r{=}0.50$ but 77% of seeds at $r{=}0.90$. The effect is absent in 1-layer models, where imbalance instead traps the model on the shortcut. Through mechanistic analysis -- gradient conflict dynamics, circuit evolution, and QK/OV circuit ablations -- we characterize a mechanistic pathway consistent with imbalance promoting generalization.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
A Unified Generalization Analysis of Re-Weighting and Logit-Adjustment for Imbalanced Learning
Real-world datasets are typically imbalanced in the sense that only a few classes have numerous samples, while many classes are associated with only a few samples. As a result, a na\"ive ERM learning process will be bias…
A step towards understanding why classification helps regression
A number of computer vision deep regression approaches report improved results when adding a classification loss to the regression loss. Here, we explore why this is useful in practice and when it is beneficial. To do so…
Age EstimationClassificationDepth EstimationregressionA Unified Generalization Analysis of Re-Weighting and Logit-Adjustment for Imbalanced Learning. paper with code
Real-world datasets are typically imbalanced in the sense that only a few classes have numerous samples, while many classes are associated with only a few samples. As a result, a naïve ERM learning process will be biased…
Long-tail LearningDP-SGD vs PATE: Which Has Less Disparate Impact on GANs?
Generative Adversarial Networks (GANs) are among the most popular approaches to generate synthetic data, especially images, for data sharing purposes. Given the vital importance of preserving the privacy of the individua…
How Re-sampling Helps for Long-Tail Learning?
Long-tail learning has received significant attention in recent years due to the challenge it poses with extremely imbalanced datasets. In these datasets, only a few classes (known as the head classes) have an adequate n…