Metraipolis › archive › dc_bfe91467985e
Exploiting answer-invariant redundancies in satellite imagery for efficient VLM inference on edge — arXiv:2609.29029
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.29029
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T05:26:27.294Z
- SHA-256 of the archived text
b80a28bcdcf00be79d595fb0925df3d5060871532639111267d0ebcbf0786001- Stable URL of this page
https://metraipolis.com/documento/dc_bfe91467985e
How to cite
Exploiting answer-invariant redundancies in satellite imagery for efficient VLM inference on edge — arXiv:2609.29029 — arXiv (en), https://arxiv.org/abs/2609.29029 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_bfe91467985e (Metraipolis archive dc_bfe91467985e, sha256 b80a28bcdcf00be79d595fb0925df3d5060871532639111267d0ebcbf0786001)
BibTeX
@misc{dcbfe91467985e,
title = {Exploiting answer-invariant redundancies in satellite imagery for efficient VLM inference on edge — arXiv:2609.29029},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.29029},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_bfe91467985e; sha256 b80a28bcdcf00be79d595fb0925df3d5060871532639111267d0ebcbf0786001},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.29029. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Onboard vision-language models could enable satellites to answer queries directly, but exhaustive tiled inference over high-resolution imagery is slow and energy-intensive. We identify answer-invariant token redundancy (AITR): image tiles and vision tokens that can be removed without changing the final answer. We present Rift, a two-stage system that performs query-conditioned tile pruning followed by elastic prefill to reduce token budget. We evaluate it on LLaVA-1.5 7B running on Jetson AGX Orin. Compared with exhaustive tiled inference, Rift reduces energy by 78% and latency by 69%, while increasing accuracy from 45% to 73%. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-25 in cs.LG). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2609.29029 — 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 2026-10-08T04:50:39.574Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-25?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).