Metraipolis › archive › dc_12405af461cd

VPRune: Efficient Training-free Pre-LLM Visual Token Pruning — arXiv:2609.24485

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.24485
License (as archived): CC0-1.0
Archived: 2026-10-09T22:05:59.972Z
SHA-256 of the archived text: 2d44f701a163eccee62fdad08837a4384a7340c29c5551e399a503daa0f26ecf
Stable URL of this page: https://metraipolis.com/documento/dc_12405af461cd

How to cite

VPRune: Efficient Training-free Pre-LLM Visual Token Pruning — arXiv:2609.24485 — arXiv (en), https://arxiv.org/abs/2609.24485 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_12405af461cd (Metraipolis archive dc_12405af461cd, sha256 2d44f701a163eccee62fdad08837a4384a7340c29c5551e399a503daa0f26ecf)

BibTeX

@misc{dc12405af461cd,
  title = {VPRune: Efficient Training-free Pre-LLM Visual Token Pruning — arXiv:2609.24485},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.24485},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_12405af461cd; sha256 2d44f701a163eccee62fdad08837a4384a7340c29c5551e399a503daa0f26ecf},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.24485. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Visual token pruning is a promising approach to reducing the inference cost of large vision-language models (LVLMs), yet aggressive token reduction often causes substantial performance degradation. We identify three key factors behind this degradation: text-guided selection bias, information loss from discarded tokens, and positional distortion caused by sequence compaction. Based on these observations, we propose \textbf{VPRune}, a training-free pre-LLM pruning framework consisting of visual-only diversity selection, similarity-guided token recycling, and position-preserving restoration. Experiments on FastVLM-1.5B across multiple vision-language benchmarks demonstrate that VPRune achieves a favorable accuracy--compression trade-off, with particularly pronounced advantages under aggressive compression. Furthermore, evaluations on edge-device show that VPRune effectively reduces end-to-end inference latency while maintaining superior task performance, demonstrating its practicality for resource-constrained LVLM deployment.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-22 in cs.CV).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.24485 — 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 2026-10-09T21:45:21.047Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-09-22?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).