# Learning Diverse Features in Vision Transformers for Improved Generalization

arXiv: https://arxiv.org/abs/2308.16274 · HTML: https://mlsanity.com/paper/2308.16274/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Builds on 25 of its 26 references that are in the map.
- First posted on arXiv 30 Aug 2023.

## Builds on (most cited first, 25 of 25)

- [Shortcut Learning in Deep Neural Networks](https://mlsanity.com/paper/2004.07780/) (2020)
- [Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalization](https://mlsanity.com/paper/1911.08731/) (2019)
- [Transformers in Vision: A Survey](https://mlsanity.com/paper/2101.01169/) (2021)
- [Underspecification Presents Challenges for Credibility in Modern Machine Learning](https://mlsanity.com/paper/2011.03395/) (2020)
- [Intriguing Properties of Vision Transformers](https://mlsanity.com/paper/2105.10497/) (2021)
- [The Pitfalls of Simplicity Bias in Neural Networks](https://mlsanity.com/paper/2006.07710/) (2020)
- [Understanding Robustness of Transformers for Image Classification](https://mlsanity.com/paper/2103.14586/) (2021)
- [Leveraging Unlabeled Data to Predict Out-of-Distribution Performance](https://mlsanity.com/paper/2201.04234/) (2022)
- [Understanding The Robustness in Vision Transformers](https://mlsanity.com/paper/2204.12451/) (2022)
- [Agreement-on-the-Line: Predicting the Performance of Neural Networks under Distribution Shift](https://mlsanity.com/paper/2206.13089/) (2022)
- [On the Existence of Simpler Machine Learning Models](https://mlsanity.com/paper/1908.01755/) (2019)
- [Evading the Simplicity Bias: Training a Diverse Set of Models Discovers Solutions with Superior OOD Generalization](https://mlsanity.com/paper/2105.05612/) (2021)
- [Agree to Disagree: Diversity through Disagreement for Better Transferability](https://mlsanity.com/paper/2202.04414/) (2022)
- [Diversify and Disambiguate: Learning From Underspecified Data](https://mlsanity.com/paper/2202.03418/) (2022)
- [Understanding and Improving Feature Learning for Out-of-Distribution Generalization](https://mlsanity.com/paper/2304.11327/) (2023)
- [Rich Feature Construction for the Optimization-Generalization Dilemma](https://mlsanity.com/paper/2203.15516/) (2022)
- [Exploring and Interacting with the Set of Good Sparse Generalized Additive Models](https://mlsanity.com/paper/2303.16047/) (2023)
- [Neuron Activation Coverage: Rethinking Out-of-distribution Detection and Generalization](https://mlsanity.com/paper/2306.02879/) (2023)
- [Predicting is not Understanding: Recognizing and Addressing Underspecification in Machine Learning](https://mlsanity.com/paper/2207.02598/) (2022)
- [Ensembles of Locally Independent Prediction Models](https://mlsanity.com/paper/1911.01291/) (2019)
- [Confidence and Dispersity Speak: Characterising Prediction Matrix for Unsupervised Accuracy Estimation](https://mlsanity.com/paper/2302.01094/) (2023)
- [Learning Qualitatively Diverse and Interpretable Rules for Classification](https://mlsanity.com/paper/1806.08716/) (2018)
- [Project and Probe: Sample-Efficient Domain Adaptation by Interpolating Orthogonal Features](https://mlsanity.com/paper/2302.05441/) (2023)
- [Feature Space Particle Inference for Neural Network Ensembles](https://mlsanity.com/paper/2206.00944/) (2022)
- [Towards Optimization and Model Selection for Domain Generalization: A Mixup-guided Solution](https://mlsanity.com/paper/2209.00652/) (2022)

## Built on by (most cited first, 4 of 4)

- [Leveraging Diffusion Disentangled Representations to Mitigate Shortcuts in Underspecified Visual Tasks](https://mlsanity.com/paper/2310.02230/) (2023)
- [Mitigating Shortcut Learning with Diffusion Counterfactuals and Diverse Ensembles](https://mlsanity.com/paper/2311.16176/) (2023)
- [Prompt-Guided Attention Head Selection for Focus-Oriented Image Retrieval](https://mlsanity.com/paper/2504.01348/) (2025)
- [What Does Attention Transfer Transfer? Attention Structure and Robustness in Vision Transformers](https://mlsanity.com/paper/2608.18399/) (2026)

## Abstract (arXiv)

Deep learning models often rely only on a small set of features even when there is a rich set of predictive signals in the training data. This makes models brittle and sensitive to distribution shifts. In this work, we first examine vision transformers (ViTs) and find that they tend to extract robust and spurious features with distinct attention heads. As a result of this modularity, their performance under distribution shifts can be significantly improved at test time by pruning heads corresponding to spurious features, which we demonstrate using an "oracle selection" on validation data. Second, we propose a method to further enhance the diversity and complementarity of the learned features by encouraging orthogonality of the attention heads' input gradients. We observe improved out-of-distribution performance on diagnostic benchmarks (MNIST-CIFAR, Waterbirds) as a consequence of the enhanced diversity of features and the pruning of undesirable heads.