Metraipolis › archive › dc_5367d423541f

Canonical Color as a Lens into Concept Decodability in Vision Encoders and VLMs — arXiv:2609.09124

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.09124
License (as archived)
CC0-1.0
Archived
2026-10-16T15:12:33.992Z
SHA-256 of the archived text
b912e49d47e81d648553945764feadf92c3b31d0e2cb9f4c52decec8808f5fc6
Stable URL of this page
https://metraipolis.com/documento/dc_5367d423541f

How to cite

Canonical Color as a Lens into Concept Decodability in Vision Encoders and VLMs — arXiv:2609.09124 — arXiv (en), https://arxiv.org/abs/2609.09124 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_5367d423541f (Metraipolis archive dc_5367d423541f, sha256 b912e49d47e81d648553945764feadf92c3b31d0e2cb9f4c52decec8808f5fc6)
BibTeX
@misc{dc5367d423541f,
  title = {Canonical Color as a Lens into Concept Decodability in Vision Encoders and VLMs — arXiv:2609.09124},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.09124},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_5367d423541f; sha256 b912e49d47e81d648553945764feadf92c3b31d0e2cb9f4c52decec8808f5fc6},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.09124. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Visual encoders construct a representation of the image input for Vision-Language models. How much conceptual, as opposed to immediately visible, information does this representation contain? We use canonical color as a controlled test case to ask whether vision encoders make canonical-color information linearly accessible, even when color is removed from the input image. We construct a dataset of objects with canonical colors, and probe vision encoders for both color and object identity using color and grayscale images. We find that canonical color remains decodable from grayscale images, and is tied to predicted object identity, indicating a conceptual link. Extending this analysis to full VLMs, we find that VLM post-training can have a surprisingly large effect on color decodability in the vision encoder. Overall, canonical color provides a usefully controllable lens for tracing object-level conceptual semantic information in vision encoders and VLMs.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-09 in cs.CV).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.09124 — 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 2026-10-16T14:49:57.397Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-09-09?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).