Metraipolis › archive › dc_55ac41f64375

Decoding the Functional Roles of Register and High-Norm Patch Tokens in Vision Transformers — arXiv:2610.03698

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.03698
License (as archived)
CC0-1.0
Archived
2026-10-06T09:09:05.712Z
SHA-256 of the archived text
9d9a8f81952b4f4a777f7105f2b9b1c6c43c52adbd37e047490603d5fa0daff1
Stable URL of this page
https://metraipolis.com/documento/dc_55ac41f64375

How to cite

Decoding the Functional Roles of Register and High-Norm Patch Tokens in Vision Transformers — arXiv:2610.03698 — arXiv (en), https://arxiv.org/abs/2610.03698 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_55ac41f64375 (Metraipolis archive dc_55ac41f64375, sha256 9d9a8f81952b4f4a777f7105f2b9b1c6c43c52adbd37e047490603d5fa0daff1)
BibTeX
@misc{dc55ac41f64375,
  title = {Decoding the Functional Roles of Register and High-Norm Patch Tokens in Vision Transformers — arXiv:2610.03698},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.03698},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_55ac41f64375; sha256 9d9a8f81952b4f4a777f7105f2b9b1c6c43c52adbd37e047490603d5fa0daff1},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.03698. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Self-supervised Vision Transformers (ViTs), such as DINOv2, learn rich visual representations, but the functions of their internal tokens remain poorly understood. Recent architectures introduce dedicated register tokens to reduce high-norm out- lier patch tokens that emerge in background re- gions, yet the semantic and functional roles of both token types have not been fully established. In this paper, we analyze these roles by training sparse autoencoders (SAEs) on register-token and outlier-token activations in DINOv2. Using an automated interpretability pipeline, UMAP clus- tering, and CLIP-space cross-checks, we find that register-token features are more strongly associ- ated with high-level semantic concepts. Outlier- token features, by contrast, are more often associ- ated with lower-level structural, background, and texture-dominant patterns. Causal ablations fur- ther reveal a substantial functional asymmetry: disrupting top-activating register-derived features produces a 48.17% drop in representation cosine similarity, whereas disrupting outlier-derived fea- tures produces only a 0.31% drop. Together, our results provide evidence for token specialization in self-supervised ViTs.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-10-05 in cs.CV).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2610.03698 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-06T08:21:44.276Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-10-05?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).