MUNIChus: Multilingual News Image Captioning Benchmark

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chen, Yuji, Plum, Alistair, Hettiarachchi, Hansi, Kanojia, Diptesh, Basnet, Saroj, Zampieri, Marcos, Ranasinghe, Tharindu
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917332873379840
author Chen, Yuji
Plum, Alistair
Hettiarachchi, Hansi
Kanojia, Diptesh
Basnet, Saroj
Zampieri, Marcos
Ranasinghe, Tharindu
author_facet Chen, Yuji
Plum, Alistair
Hettiarachchi, Hansi
Kanojia, Diptesh
Basnet, Saroj
Zampieri, Marcos
Ranasinghe, Tharindu
contents The goal of news image captioning is to generate captions by integrating news article content with corresponding images, highlighting the relationship between textual context and visual elements. The majority of research on news image captioning focuses on English, primarily because datasets in other languages are scarce. To address this limitation, we create the first multilingual news image captioning benchmark, MUNIChus, comprising 9 languages, including several low-resource languages such as Sinhala and Urdu. We evaluate various state-of-the-art neural news image captioning models on MUNIChus and find that news image captioning remains challenging. We also make MUNIChus publicly available with over 20 models already benchmarked. MUNIChus opens new avenues for further advancements in developing and evaluating multilingual news image captioning models.
format Preprint
id arxiv_https___arxiv_org_abs_2603_10613
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle MUNIChus: Multilingual News Image Captioning Benchmark
Chen, Yuji
Plum, Alistair
Hettiarachchi, Hansi
Kanojia, Diptesh
Basnet, Saroj
Zampieri, Marcos
Ranasinghe, Tharindu
Computation and Language
Computer Vision and Pattern Recognition
The goal of news image captioning is to generate captions by integrating news article content with corresponding images, highlighting the relationship between textual context and visual elements. The majority of research on news image captioning focuses on English, primarily because datasets in other languages are scarce. To address this limitation, we create the first multilingual news image captioning benchmark, MUNIChus, comprising 9 languages, including several low-resource languages such as Sinhala and Urdu. We evaluate various state-of-the-art neural news image captioning models on MUNIChus and find that news image captioning remains challenging. We also make MUNIChus publicly available with over 20 models already benchmarked. MUNIChus opens new avenues for further advancements in developing and evaluating multilingual news image captioning models.
title MUNIChus: Multilingual News Image Captioning Benchmark
topic Computation and Language
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.10613