Metraipolis › archive › dc_0c742942e8ed
TReVS: Integrating Textual Relevance and Visual Saliency for Efficient Vision-Language Model Token Pruning — arXiv:2609.37581
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.37581
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T16:27:24.137Z
- SHA-256 of the archived text
36c87128e621e282b3bbe5982ed5635bab932388b72f71e0f3154a3d0c0e84d1- Stable URL of this page
https://metraipolis.com/documento/dc_0c742942e8ed
How to cite
TReVS: Integrating Textual Relevance and Visual Saliency for Efficient Vision-Language Model Token Pruning — arXiv:2609.37581 — arXiv (en), https://arxiv.org/abs/2609.37581 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0c742942e8ed (Metraipolis archive dc_0c742942e8ed, sha256 36c87128e621e282b3bbe5982ed5635bab932388b72f71e0f3154a3d0c0e84d1)
BibTeX
@misc{dc0c742942e8ed,
title = {TReVS: Integrating Textual Relevance and Visual Saliency for Efficient Vision-Language Model Token Pruning — arXiv:2609.37581},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.37581},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0c742942e8ed; sha256 36c87128e621e282b3bbe5982ed5635bab932388b72f71e0f3154a3d0c0e84d1},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.37581. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Vision-Language Models (VLMs) excel at visual understanding and reasoning but often incur substantial inference costs due to the large number of visual tokens. Recent visual token pruning methods increasingly follow a two-stage paradigm: they first remove visually redundant tokens after the vision encoder and then discard tokens irrelevant to the textual query within the Large Language Model (LLM). However, since the first stage typically relies solely on vision-encoder saliency, it may prematurely eliminate query-relevant tokens, depriving the subsequent text-guided stage of critical visual evidence. Our empirical analysis shows that incorporating query guidance into first-stage pruning better preserves task-relevant evidence and consistently improves performance over vision-only saliency-based pruning. We further find that high-variance attention heads are more sensitive to the textual query and yield more discriminative text-to-vision attention signals for second-stage pruning. Motivated by these findings, we propose TReVS, a training-free framework that combines textual relevance with vision-encoder saliency for pre-LLM pruning and leverages high-variance attention heads to remove task-irrelevant tokens at shallow-to-intermediate layers of the LLM. On LLaVA-1.5-7B, TReVS retains 92.8% of the unpruned baseline performance while pruning 94.4% of visual tokens, outperforming prior state-of-the-art methods. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-30 in cs.AI). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.37581 — 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 2026-10-07T15:54:53.995Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).