Metraipolis › archive › dc_3ff4d124cae5

MiX: Micro-Inverted-Scaling for End-to-End Low-Bit Vision-Language Model Acceleration — arXiv:2609.19683

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.19683
License (as archived)
CC0-1.0
Archived
2026-10-10T03:51:23.011Z
SHA-256 of the archived text
fd2629dd22d7fe060b1e5dd4d34cf8b3a280ead6b726b125715a9c8df0ea0271
Stable URL of this page
https://metraipolis.com/documento/dc_3ff4d124cae5

How to cite

MiX: Micro-Inverted-Scaling for End-to-End Low-Bit Vision-Language Model Acceleration — arXiv:2609.19683 — arXiv (en), https://arxiv.org/abs/2609.19683 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_3ff4d124cae5 (Metraipolis archive dc_3ff4d124cae5, sha256 fd2629dd22d7fe060b1e5dd4d34cf8b3a280ead6b726b125715a9c8df0ea0271)
BibTeX
@misc{dc3ff4d124cae5,
  title = {MiX: Micro-Inverted-Scaling for End-to-End Low-Bit Vision-Language Model Acceleration — arXiv:2609.19683},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.19683},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_3ff4d124cae5; sha256 fd2629dd22d7fe060b1e5dd4d34cf8b3a280ead6b726b125715a9c8df0ea0271},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.19683. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

The deployment of Vision-Language Models (VLMs) on edge devices is severely bottlenecked by memory bandwidth, necessitating aggressive sub-8-bit quantization. Since edge accelerators are strictly constrained by area and power, they require end-to-end quantized models. However, the extreme dynamic range gap between multi-modal tokens causes standard block formats to suffer "microscaling collapse," where a single massive outlier hijacks the shared exponent, underflowing surrounding elements and destroying attention maps. To break this bottleneck, we propose Micro-Inverted-Scaling (MiX), a novel format that mathematically inverts the microscaling paradigm: rather than grouping multiple mantissas under one shared exponent, MiX groups private, per-element exponents under a single shared mantissa. To handle asymmetric VLM outlier topologies, we introduce an adaptive dual-format (MiX-MX) inference framework. By algebraically factoring out the shared MiX mantissa, this framework maps to a custom accelerator, replacing multipliers with efficient shifters. Evaluated end-to-end on multiple VLMs, our 4.5-bit MiX formulation exhibits equivalent or superior accuracy on multi-modal benchmarks compared to NVFP4. Simultaneously, the MiX accelerator delivers a 25% improvement in area efficiency over the NVFP4 baseline and a 2.3-4.5x speedup with 1.4-2.9x energy reduction across models compared to the state-of-the-art accelerator Focus, proving the inverted-scaling datapath is physically superior for efficient VLM deployment.

Tema degli esploratori: calcolo e infrastrutture (retro-catalogo arXiv: annunci del 2026-09-18 in cs.AR).
Categorie arXiv: Hardware Architecture (cs.AR) ; Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2609.19683 — 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 2026-10-10T03:24:33.088Z dalla pagina https://arxiv.org/catchup/cs.AR/2026-09-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).