Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

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

About

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.

Hiroyuki Deguchi, Katsuki Chousa, Yusuke Sakai• 2026

Related benchmarks

TaskDatasetResultRank
Text-to-Image AlignmentMS-COCO
CLIP Score1.023
60
Image-Text Alignment EvaluationNoCaps Out-of-Domain (val)
CLIPScore84.1
40
Caption EvaluationMSCOCO
Win Rate90
20
Caption EvaluationNoCaps
Win Rate71.1
20
Image CaptioningMSCOCO
CLIPScore40.6
14
Showing 5 of 5 rows

Other info

Follow for update