Metraipolis › archive › dc_08fc76aa9218

More Than Seeing: Vision as Scalable Anchors in Native Multimodal Pre-Training — arXiv:2610.19443

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.19443
License (as archived): CC0-1.0
Archived: 2026-10-15T20:23:52.242Z
SHA-256 of the archived text: ed06522e714aa0d1c804da677e024eb2b6843e829b50313bd6ccaa522cc95bda
Stable URL of this page: https://metraipolis.com/documento/dc_08fc76aa9218

How to cite

More Than Seeing: Vision as Scalable Anchors in Native Multimodal Pre-Training — arXiv:2610.19443 — arXiv (en), https://arxiv.org/abs/2610.19443 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_08fc76aa9218 (Metraipolis archive dc_08fc76aa9218, sha256 ed06522e714aa0d1c804da677e024eb2b6843e829b50313bd6ccaa522cc95bda)

BibTeX

@misc{dc08fc76aa9218,
  title = {More Than Seeing: Vision as Scalable Anchors in Native Multimodal Pre-Training — arXiv:2610.19443},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.19443},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_08fc76aa9218; sha256 ed06522e714aa0d1c804da677e024eb2b6843e829b50313bd6ccaa522cc95bda},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.19443. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Native multimodal pre-training trains a single model from scratch on text and multimodal data, enabling it to understand both language and visual inputs. However, how visual signals shape scaling behavior and support capabilities beyond visual understanding remains unclear. We investigate this question through controlled visual-content ablations across model sizes and training budgets. Specifically, we replace image embeddings with content-free padding token embeddings. Retaining visual content improves text prediction on image-text examples without materially changing the lowest text-only loss at a given compute budget, the optimal balance of model size and training data, or aggregate language scores. The relative multimodal loss benefit grows approximately as a power law of compute. To test benefits beyond understanding images, we evaluate spatial reasoning from text descriptions alone. Both conditions improve with model and data scaling. During training of the largest model, visual gains are most sustained in localization, navigation, and symbolic planning, which involve maintaining object configurations or composing spatial updates; gains in spatial relations, geometry, and physical commonsense are less consistent. These findings identify visual content as a scalable anchor for multimodal learning, with benefits that extend beyond visual inputs while preserving aggregate language performance.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.19443 — 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 2026-10-15T19:36:18.192Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).