One Single Hub Text Breaks CLIP: Identifying Vulnerabilities in Cross-Modal Encoders via Hubness

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Deguchi, Hiroyuki, Chousa, Katsuki, Sakai, Yusuke
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911636055392256
author Deguchi, Hiroyuki
Chousa, Katsuki
Sakai, Yusuke
author_facet Deguchi, Hiroyuki
Chousa, Katsuki
Sakai, Yusuke
contents The hubness problem, in which hub embeddings are close to many unrelated examples, occurs often in high-dimensional embedding spaces and may pose a practical threat for purposes such as information retrieval and automatic evaluation metrics. In particular, since cross-modal similarity between text and images cannot be calculated by direct comparisons, such as string matching, cross-modal encoders that project different modalities into a shared space are helpful for various cross-modal applications, and thus, the existence of hubs may pose practical threats. To reveal the vulnerabilities of cross-modal encoders, we propose a method for identifying the hub embedding and its corresponding hub text. Experiments on image captioning evaluation in MSCOCO and nocaps along with image-to-text retrieval tasks in MSCOCO and Flickr30k showed that our method can identify a single hub text that unreasonably achieves comparable or higher similarity scores than human-written reference captions in many images, thereby revealing the vulnerabilities in cross-modal encoders.
format Preprint
id arxiv_https___arxiv_org_abs_2604_27674
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle One Single Hub Text Breaks CLIP: Identifying Vulnerabilities in Cross-Modal Encoders via Hubness
Deguchi, Hiroyuki
Chousa, Katsuki
Sakai, Yusuke
Computation and Language
Artificial Intelligence
Cryptography and Security
Information Retrieval
The hubness problem, in which hub embeddings are close to many unrelated examples, occurs often in high-dimensional embedding spaces and may pose a practical threat for purposes such as information retrieval and automatic evaluation metrics. In particular, since cross-modal similarity between text and images cannot be calculated by direct comparisons, such as string matching, cross-modal encoders that project different modalities into a shared space are helpful for various cross-modal applications, and thus, the existence of hubs may pose practical threats. To reveal the vulnerabilities of cross-modal encoders, we propose a method for identifying the hub embedding and its corresponding hub text. Experiments on image captioning evaluation in MSCOCO and nocaps along with image-to-text retrieval tasks in MSCOCO and Flickr30k showed that our method can identify a single hub text that unreasonably achieves comparable or higher similarity scores than human-written reference captions in many images, thereby revealing the vulnerabilities in cross-modal encoders.
title One Single Hub Text Breaks CLIP: Identifying Vulnerabilities in Cross-Modal Encoders via Hubness
topic Computation and Language
Artificial Intelligence
Cryptography and Security
Information Retrieval
url https://arxiv.org/abs/2604.27674