Metraipolis › archive › dc_22d298b27bbd
Inverting Multi-Vector Visual Document Indices — arXiv:2610.09920
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.09920
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T11:26:37.732Z
- SHA-256 of the archived text
99a2a5d27a30fae0879cd41235dadf47540b7e34dc652073b55469f522968cbf- Stable URL of this page
https://metraipolis.com/documento/dc_22d298b27bbd
How to cite
Inverting Multi-Vector Visual Document Indices — arXiv:2610.09920 — arXiv (en), https://arxiv.org/abs/2610.09920 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_22d298b27bbd (Metraipolis archive dc_22d298b27bbd, sha256 99a2a5d27a30fae0879cd41235dadf47540b7e34dc652073b55469f522968cbf)
BibTeX
@misc{dc22d298b27bbd,
title = {Inverting Multi-Vector Visual Document Indices — arXiv:2610.09920},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.09920},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_22d298b27bbd; sha256 99a2a5d27a30fae0879cd41235dadf47540b7e34dc652073b55469f522968cbf},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.09920. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Prevailing multi-vector visual document retrievers store each page as about a thousand patch vectors, often in vector databases run by a third party. Since no one can read a page from its vectors, this index is easily treated as less sensitive than the page. However, because the index keeps one vector per patch in raster order, and each vector is computed by a vision-language model pre-trained to read documents, we hypothesize that whoever runs or breaches the store can reproduce a page from its index alone. We frame inversion as conditional document image generation and infer from the vectors what the attack needs: the encoder, the page shape and, for shuffled vectors, their order. On the ViDoRe v3 benchmark, pages inverted from raw indices recover 47% of the words and 45% of the sensitive tokens. Used as queries against the stored indices, they rank their source page first 98.4% of the time. We test two cheap protections, token pooling and shuffling, which both cut word recall to about 8%. A model that restores the order of a shuffled index raises the share of source pages ranked first from 3.8% to 93.5%, while inverting a pooled index remains open. To test generalisation, we apply the same attack unchanged to another multi-vector retriever: its inverted pages still rank their source page first 70.2% of the time, though its word recall stays below a nearest-neighbour baseline. Multi-vector visual document retrievers are therefore vulnerable to inversion through their stored index, which should be protected like the documents it encodes. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno). Categorie arXiv: Information Retrieval (cs.IR) ; Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2610.09920 — 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 2026-10-08T10:24:06.388Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).