Maximal Matching Matters: Preventing Representation Collapse for Robust Cross-Modal Retrieval

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Alomari, Hani, Sivakumar, Anushka, Zhang, Andrew, Thomas, Chris
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866915360172670976
author Alomari, Hani
Sivakumar, Anushka
Zhang, Andrew
Thomas, Chris
author_facet Alomari, Hani
Sivakumar, Anushka
Zhang, Andrew
Thomas, Chris
contents Cross-modal image-text retrieval is challenging because of the diverse possible associations between content from different modalities. Traditional methods learn a single-vector embedding to represent semantics of each sample, but struggle to capture nuanced and diverse relationships that can exist across modalities. Set-based approaches, which represent each sample with multiple embeddings, offer a promising alternative, as they can capture richer and more diverse relationships. In this paper, we show that, despite their promise, these set-based representations continue to face issues including sparse supervision and set collapse, which limits their effectiveness. To address these challenges, we propose Maximal Pair Assignment Similarity to optimize one-to-one matching between embedding sets which preserve semantic diversity within the set. We also introduce two loss functions to further enhance the representations: Global Discriminative Loss to enhance distinction among embeddings, and Intra-Set Divergence Loss to prevent collapse within each set. Our method achieves state-of-the-art performance on MS-COCO and Flickr30k without relying on external data.
format Preprint
id arxiv_https___arxiv_org_abs_2506_21538
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Maximal Matching Matters: Preventing Representation Collapse for Robust Cross-Modal Retrieval
Alomari, Hani
Sivakumar, Anushka
Zhang, Andrew
Thomas, Chris
Computer Vision and Pattern Recognition
Information Retrieval
Machine Learning
Cross-modal image-text retrieval is challenging because of the diverse possible associations between content from different modalities. Traditional methods learn a single-vector embedding to represent semantics of each sample, but struggle to capture nuanced and diverse relationships that can exist across modalities. Set-based approaches, which represent each sample with multiple embeddings, offer a promising alternative, as they can capture richer and more diverse relationships. In this paper, we show that, despite their promise, these set-based representations continue to face issues including sparse supervision and set collapse, which limits their effectiveness. To address these challenges, we propose Maximal Pair Assignment Similarity to optimize one-to-one matching between embedding sets which preserve semantic diversity within the set. We also introduce two loss functions to further enhance the representations: Global Discriminative Loss to enhance distinction among embeddings, and Intra-Set Divergence Loss to prevent collapse within each set. Our method achieves state-of-the-art performance on MS-COCO and Flickr30k without relying on external data.
title Maximal Matching Matters: Preventing Representation Collapse for Robust Cross-Modal Retrieval
topic Computer Vision and Pattern Recognition
Information Retrieval
Machine Learning
url https://arxiv.org/abs/2506.21538