Metraipolis › archive › dc_f2ed7d8b5d52

VisionWeave: Weaving Elastic Visual Representations as a Native Capability of MLLMs — arXiv:2610.07987

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.07987
License (as archived)
CC0-1.0
Archived
2026-10-07T11:21:25.452Z
SHA-256 of the archived text
996bbaf93a9e61f51ddfb35ed8f0a0c75ee50bd9e8a8f712d03d91f0a7565e77
Stable URL of this page
https://metraipolis.com/documento/dc_f2ed7d8b5d52

How to cite

VisionWeave: Weaving Elastic Visual Representations as a Native Capability of MLLMs — arXiv:2610.07987 — arXiv (en), https://arxiv.org/abs/2610.07987 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f2ed7d8b5d52 (Metraipolis archive dc_f2ed7d8b5d52, sha256 996bbaf93a9e61f51ddfb35ed8f0a0c75ee50bd9e8a8f712d03d91f0a7565e77)
BibTeX
@misc{dcf2ed7d8b5d52,
  title = {VisionWeave: Weaving Elastic Visual Representations as a Native Capability of MLLMs — arXiv:2610.07987},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.07987},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f2ed7d8b5d52; sha256 996bbaf93a9e61f51ddfb35ed8f0a0c75ee50bd9e8a8f712d03d91f0a7565e77},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.07987. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Multimodal large language models have become the dominant paradigm for visual understanding, but incur substantial costs by encoding inputs into dense, fixed-size patch tokens. However, visual information is unevenly distributed: some regions require fine-grained detail, while others admit compact representations. Downsampling sacrifices this detail, while existing token pruning and adaptive approaches remain limited in content-adaptive granularity, task generalization, and integration with modern MLLMs and serving infrastructure. Overcoming these limitations calls for foundation models that learn, end to end, where-and at what granularity-to allocate visual representations, a native capability we term elastic visual representation weaving. We introduce VisionWeave, establishing this capability in frontier-level MLLMs through large-scale training. It combines two components: a gated spatial pooler constructs coarse-grained representations alongside native fine-grained representations within a shared MRoPE coordinate, while a granularity router learns their content-adaptive allocation. Through self-distillation alone, we validate this capability on Qwen3.5-4B and scale to Qwen3.8-27B with over 30K A100 GPU-hours. Based on Qwen3.8-27B, VisionWeave adaptively adjusts token savings to visual content, saving 43.0% tokens on average while retaining 98.9% native performance across eight benchmarks, versus only 88% performance preserved for token pruning baselines with a fixed 50% savings target. Extensive evaluations confirm robust efficiency-quality trade-offs across diverse tasks, resolutions and video frames. When deployed on SGLang serving engine, our method achieves a 2.3x throughput gain while reducing mean TTFT by 54.4% and mean TPOT by 60.6%. Together, we believe these results position elastic visual weaving as a promising capability for next-generation multimodal models.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.07987 — 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 2026-10-07T11:15:24.791Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).