English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Multimodal Machine Translation (MMT) aims to improve translation quality byleveraging auxiliary modalities such as images alongside textual input. Whilerecent advances in large-scale pre-trained language and vision models havesignificantly benefited unimodal natural language processing tasks, theireffectiveness and role in MMT remain underexplored. In this work, we conduct asystematic study on the impact of pre-trained encoders and decoders inmultimodal translation models. Specifically, we analyze how different trainingstrategies, from training from scratch to using pre-trained and partiallyfrozen components, affect translation performance under a unified MMTframework. Experiments are carried out on the Multi30K and CoMMuTE datasetacross English-German and English-French translation tasks. Our results revealthat pre-training plays a crucial yet asymmetrical role in multimodal settings:pre-trained decoders consistently yield more fluent and accurate outputs, whilepre-trained encoders show varied effects depending on the quality ofvisual-text alignment. Furthermore, we provide insights into the interplaybetween modality fusion and pre-trained components, offering guidance forfuture architecture design in multimodal translation systems.

Memory Reviving, Continuing Learning and Beyond: Evaluation of Pre-trained Encoders and Decoders for Multimodal Machine Translation