Metraipolis › archive › dc_3faeb21efc4a

Adaptive Visual Token Reduction for Accelerated Image Understanding — arXiv:2610.09252

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.09252
License (as archived)
CC0-1.0
Archived
2026-10-08T12:47:03.093Z
SHA-256 of the archived text
8a3141556eb39af9589cbe3bf9cda8891e47c43804d8e1328169795b22470f6c
Stable URL of this page
https://metraipolis.com/documento/dc_3faeb21efc4a

How to cite

Adaptive Visual Token Reduction for Accelerated Image Understanding — arXiv:2610.09252 — arXiv (en), https://arxiv.org/abs/2610.09252 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_3faeb21efc4a (Metraipolis archive dc_3faeb21efc4a, sha256 8a3141556eb39af9589cbe3bf9cda8891e47c43804d8e1328169795b22470f6c)
BibTeX
@misc{dc3faeb21efc4a,
  title = {Adaptive Visual Token Reduction for Accelerated Image Understanding — arXiv:2610.09252},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.09252},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_3faeb21efc4a; sha256 8a3141556eb39af9589cbe3bf9cda8891e47c43804d8e1328169795b22470f6c},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.09252. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large Vision-Language Models achieve strong VQA performance, but processing high-resolution, information-rich images requires substantial computation, motivating visual token reduction. However, existing methods often prune individual tokens or rely on fixed-size cropping, limiting their ability to preserve spatially structured information such as horizontally or vertically elongated text. To address this limitation, we propose ReFIT, an instruction-guided visual token reduction framework for efficient LVLM inference. ReFIT consists of Relevance-Guided Window Reshaping (RWR) and Instruction-Guided Token Refinement (ITR), where RWR captures instruction-relevant regions by adapting to their spatial characteristics, while ITR further removes unnecessary visual tokens. Experiments on four VQA benchmarks demonstrate that ReFIT improves answer accuracy while reducing computational cost, and qualitative results demonstrate its effectiveness in localizing relevant regions and removing unnecessary visual information.

Tema degli esploratori: statistica e scienza dei dati (elenco arXiv cs.CV, nuovi lavori del giorno).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2610.09252 — 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 2026-10-08T12:18:15.322Z dalla pagina https://arxiv.org/list/cs.CV/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).