Explanation is All You Need in Distillation: Mitigating Bias and Shortcut Learning
Bias and spurious correlations in data can cause shortcut learning, undermining out-of-distribution (OOD) generalization in deep neural networks. Most methods require unbiased data during training (and/or hyper-parameter tuning) to counteract shortcut learning. Here, we propose the use of explanation distillation to hinder shortcut learning. The technique does not assume any access to unbiased data, and it allows an arbitrarily sized student network to learn the reasons behind the decisions of an unbiased teacher, such as a vision-language model or a network processing debiased images. We found that it is possible to train a neural network with explanation (e.g by Layer Relevance Propagation, LRP) distillation only, and that the technique leads to high resistance to shortcut learning, surpassing group-invariant learning, explanation background minimization, and alternative distillation techniques. In the COLOURED MNIST dataset, LRP distillation achieved 98.2% OOD accuracy, while deep feature distillation and IRM achieved 92.1% and 60.2%, respectively. In COCO-on-Places, the undesirable generalization gap between in-distribution and OOD accuracy is only of 4.4% for LRP distillation, while the other two techniques present gaps of 15.1% and 52.1%, respectively.
Code (0)
등록된 구현이 없습니다.
Tasks
AllLanguage ModelingLanguage ModellingSimilar Papers 제목 키워드 기반
Don't blame Dataset Shift! Shortcut Learning due to Gradients and Cross Entropy
Common explanations for shortcut learning assume that the shortcut improves prediction under the training distribution but not in the test distribution. Thus, models trained via the typical gradient-based optimization of…
Inductive BiasDon’t blame Dataset Shift! Shortcut Learning due to Gradients and Cross Entropy
Common explanations for shortcut learning assume that the shortcut improves prediction only under the training distribution. Thus, models trained in the typical way by minimizing log-loss using gradient descent, which we…
SCISSOR: Mitigating Semantic Bias through Cluster-Aware Siamese Networks for Robust Classification
Shortcut learning undermines model generalization to out-of-distribution data. While the literature attributes shortcuts to biases in superficial features, we show that imbalances in the semantic distribution of sample e…
Data AugmentationRobust classificationLast to Learn Bias: Analyzing and Mitigating a Shortcut in Question Matching
Recent studies report that even if deep neural models make correct predictions, models may be relying on shortcut rather than understanding the semantics of the text. Previous studies indicate that the shortcut deriving …
Do LLMs Overcome Shortcut Learning? An Evaluation of Shortcut Challenges in Large Language Models
Large Language Models (LLMs) have shown remarkable capabilities in various natural language processing tasks. However, LLMs may rely on dataset biases as shortcuts for prediction, which can significantly impair their rob…
In-Context Learning