拆解 Vision Transformer 的表征结构：一项严谨的架构研究

View PDF HTML (experimental)

Abstract:Representation structure is crucial for understanding Vision Transformer (ViT) architectures and their generalization behavior. However, prior studies neither isolate nor analyze module-level features nor investigate how their interactions contribute to performance estimation. In this work, we conduct the first rigorous analysis of feature information across diverse architectural scales, empirically uncover the relationship between ViT representation and generalization behavior, and leverage these insights to guide efficient ViT design. Our contributions are fivefold: Across diverse architectural scales, 1) We identify feature collapse at initialization, which leads to redundancy, and propose a reduction scheme to mitigate this issue. 2) We quantify feature information using entropy and the minimum eigenvalue, demonstrating that these metrics serve as reliable indicators for generalization prediction. 3) We show that feature in the token space provides a more faithful representation than those in embedding space. 4) We discover an unexpected finding: features produced by linear submodules within ViT layers are critical for the prediction of generalization performance. 5) Our proposed proxy improves the correlation ranking by 18-48% over prior baselines and can effectively identify ViT architectures that achieve higher accuracy at lower or comparable computational cost.

Comments:	Accepted in IEEE VCIP 2026
Subjects:	Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
Cite as:	arXiv:2610.11205 [cs.CV]
	(or arXiv:2610.11205v1 [cs.CV] for this version)
	https://doi.org/10.48550/arXiv.2610.11205 arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Kim-Cuc Nguyen [view email]
[v1] Thu, 8 Oct 2026 04:06:50 UTC (531 KB)