The paper presents a detailed study of feature representations in Vision Transformers (ViTs), revealing that feature collapse at initialization causes redundancy and proposing a reduction scheme to address it. It demonstrates that entropy and minimum eigenvalue metrics reliably predict generalization, and that token‑space features and linear submodule outputs are particularly informative. The authors develop a proxy that improves ranking of ViT architectures by 18‑48% over previous baselines, enabling the selection of models with higher accuracy at comparable or lower computational cost.
By Kim-Cuc Nguyen, Ngai-Man Cheung
arXiv:2608. 20134v1 Announce Type: cross Abstract: We present a novel view on feature evolution in Vision Transformers (ViTs) by visualizing the training process over two dimensions -- network depth (layer) and training time (epochs).
By Joonas J\"arve, Halil Ibrahim Aysel, Tarun Khajuria, Meelis Kull
arXiv:2606. 06664v1 Announce Type: cross Abstract: Despite high accuracy, Vision Transformer (ViT) predictions can be driven by spurious cues, raising the need to understand their inner workings before safe deployment.
By Tang Li, Yanlin Chen, Mengmeng Ma, Xi Peng
arXiv:2611.05522v1 Announce Type: new
Abstract: Sparse autoencoders (SAEs) decompose visual representations into individual features, offering a promising basis for explaining model behavior. Interpr...
By Jianyang Gu, Jacob Beattie, Mridul Khurana, Julia Chae, Xi Xiao, Cheng Zhang, Anuj Karpatne, Wei-Lun Chao, Sara Beery, Tanya Berger-Wolf
The paper revisits feature inversion via inverse networks, proposing a modular variant that is computationally efficient and can produce more semantically coherent image reconstructions. It applies this method to large‑scale transformer‑based vision models—Detection Transformer, Vision Transformer, Swin Transformer, and Data‑Efficient Image Transformer—examining reconstructions across network depth. The analysis highlights gradual representational changes in Detection and Vision Transformers, systematic differences in preserving contextual shape and fine‑grained details, and varying robustness to color perturbations, while also demonstrating the framework’s broader applicability to other transformer architectures.
By Jan Rathjens, Shirin Reyhanian, David Kappel, Laurenz Wiskott
arXiv:2610.03698v1 Announce Type: new
Abstract: Self-supervised Vision Transformers (ViTs), such as DINOv2, learn rich visual representations, but the functions of their internal tokens remain poorly...
By Neel Varma, Andrew Rufail, Dipika Khullar, Vasu Sharma
The paper introduces ResLRP, an extension of Layer-wise Relevance Propagation that explicitly handles residual connections in Vision Transformers to prevent attribution explosion. It demonstrates that residual cancellation causes instability in ViT explanations, and that ResLRP improves faithfulness and localization across a wide range of ViT architectures, including Vision Language Models. The method also provides a diagnostic measure for predicting attribution degradation and successfully localizes Sparse Autoencoder features.
By Jim Berend, Reduan Achtibat, Daniel Sch\"affer, Alexander Binder, Wojciech Samek, Sebastian Lapuschkin, Maximilian Dreyer
arXiv:2607. 00434v1 Announce Type: cross Abstract: Vision-language models (VLMs) have become a paradigm for multimodal learning, yet remain unstable due to object hallucination, weak visual grounding, and catastrophic forgetting after full-parameter instruction tuning.
By Guohao Sun, Xiaofang Wang, Yash Patel, Mengchen Liu, Zhiqiang Tao, Praveen Krishnan
arXiv:2605.20337v2 Announce Type: replace
Abstract: How interpretable are the features of leading vision models? The question is increasingly pressing as these models move from research benchmarks in...
By Julien Colin, Lore Goetschalckx, Nuria Oliver, Thomas Serre
arXiv:2602.06652v2 Announce Type: replace
Abstract: The robustness of Vision Language Models (VLMs) is commonly assessed through output-level invariance, implicitly assuming that stable predictions r...
By Farooq Ahmad Wani, Alessandro Suglia, Rohit Saxena, Aryo Pradipta Gema, Wai-Chung Kwan, Fazl Barez, Maria Sofia Bucarelli, Fabrizio Silvestri, Pasquale Minervini
arXiv:2608. 14922v1 Announce Type: cross Abstract: Mechanistic interpretability has recently expanded to Vision Transformers (ViTs), with Sparse Autoencoders (SAEs) increasingly used as post-hoc tools to decompose internal representations into sparse and more interpretable features.
By Philip H. Lee, Parth Padalkar
The paper introduces Prune-and-Merge ViT, a training‑free token compression framework for Vision Transformers that combines class‑attention pruning with similarity‑based merging to reduce computational cost while preserving semantic information. It evaluates robustness using two new metrics—Relative Accuracy Drop (RAD) and Prediction Flip Rate (PFR)—and shows that the method maintains or improves accuracy on pretrained ViT and DeiT models compared to existing token reduction techniques. The authors also argue that the hybrid approach does not induce hallucination‑like behavior and keeps predictions stable under common image transformations.
By Vatsal, Soumil Malhotra, Priyanshu Chahal, Sanjeev Kumar