# When Does Pruning Benefit Vision Representations?

arXiv: https://arxiv.org/abs/2507.01722 · HTML: https://mlsanity.com/paper/2507.01722/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Builds on 20 of its 29 references that are in the map.
- First posted on arXiv 2 Jul 2025.

## Builds on (most cited first, 20 of 20)

- [Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization](https://mlsanity.com/paper/1610.02391/) (2016)
- [Axiomatic Attribution for Deep Networks](https://mlsanity.com/paper/1703.01365/) (2017)
- [The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks](https://mlsanity.com/paper/1803.03635/) (2018)
- [ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness](https://mlsanity.com/paper/1811.12231/) (2018)
- [On the Spectral Bias of Neural Networks](https://mlsanity.com/paper/1806.08734/) (2018)
- [Striving for Simplicity: The All Convolutional Net](https://mlsanity.com/paper/1412.6806/) (2014)
- [Cross-Modal Self-Attention Network for Referring Image Segmentation](https://mlsanity.com/paper/1904.04745/) (2019)
- [Comparing Rewinding and Fine-tuning in Neural Network Pruning](https://mlsanity.com/paper/2003.02389/) (2020)
- [Localizing Objects with Self-Supervised Transformers and no Labels](https://mlsanity.com/paper/2109.14279/) (2021)
- [Partial success in closing the gap between human and machine vision](https://mlsanity.com/paper/2106.07411/) (2021)
- [Ground Truth Evaluation of Neural Network Explanations with CLEVR-XAI](https://mlsanity.com/paper/2003.07258/) (2020)
- [Pruning by Explaining: A Novel Criterion for Deep Neural Network Pruning](https://mlsanity.com/paper/1912.08881/) (2019)
- [Leveraging Sparse Linear Layers for Debuggable Deep Networks](https://mlsanity.com/paper/2105.04857/) (2021)
- [Can Unstructured Pruning Reduce the Depth in Deep Neural Networks?](https://mlsanity.com/paper/2308.06619/) (2023)
- [When Does Perceptual Alignment Benefit Vision Representations?](https://mlsanity.com/paper/2410.10817/) (2024)
- [Interpreting Convolutional Neural Networks Through Compression](https://mlsanity.com/paper/1711.02329/) (2017)
- [Less is More: The Influence of Pruning on the Explainability of CNNs](https://mlsanity.com/paper/2302.08878/) (2023)
- [Towards Efficient Capsule Networks](https://mlsanity.com/paper/2208.09203/) (2022)
- [Exploring the Lottery Ticket Hypothesis with Explainability Methods: Insights into Sparse Network Performance](https://mlsanity.com/paper/2307.13698/) (2023)
- [Investigating the Effect of Network Pruning on Performance and Interpretability](https://mlsanity.com/paper/2409.19727/) (2024)

## Built on by (most cited first, 3 of 3)

- [Explainability-Inspired Layer-Wise Pruning of Deep Neural Networks for Efficient Object Detection](https://mlsanity.com/paper/2602.14040/) (2026)
- [CutClean: Neural Network Pruning for Privacy-Preserving Inference](https://mlsanity.com/paper/2608.13773/) (2026)
- [Transferring Visual Explanations: How Cross-Architecture Knowledge Distillation Affects Model Interpretability](https://mlsanity.com/paper/2609.23561/) (2026)

## Abstract (arXiv)

Pruning is widely used to reduce the complexity of deep learning models, but its effects on interpretability and representation learning remain poorly understood. This paper investigates how pruning influences vision models across three key dimensions: (i) interpretability, (ii) unsupervised object discovery, and (iii) alignment with human perception. We first analyze different vision network architectures to examine how varying sparsity levels affect feature attribution interpretability methods. Additionally, we explore whether pruning promotes more succinct and structured representations, potentially improving unsupervised object discovery by discarding redundant information while preserving essential features. Finally, we assess whether pruning enhances the alignment between model representations and human perception, investigating whether sparser models focus on more discriminative features similarly to humans. Our findings also reveal the presence of sweet spots, where sparse models exhibit higher interpretability, downstream generalization and human alignment. However, these spots highly depend on the network architectures and their size in terms of trainable parameters. Our results suggest a complex interplay between these three dimensions, highlighting the importance of investigating when and how pruning benefits vision representations.