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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Recent advances in vision-and-language modeling have seen the development ofTransformer architectures that achieve remarkable performance on multimodalreasoning tasks. Yet, the exact capabilities of these black-box models arestill poorly understood. While much of previous work has focused on studyingtheir ability to learn meaning at the word-level, their ability to tracksyntactic dependencies between words has received less attention. We take afirst step in closing this gap by creating a new multimodal task targeted atevaluating understanding of predicate-noun dependencies in a controlled setup.We evaluate a range of state-of-the-art models and find that their performanceon the task varies considerably, with some models performing relatively welland others at chance level. In an effort to explain this variability, ouranalyses indicate that the quality (and not only sheer quantity) of pretrainingdata is essential. Additionally, the best performing models leveragefine-grained multimodal pretraining objectives in addition to the standardimage-text matching objectives. This study highlights that targeted andcontrolled evaluations are a crucial step for a precise and rigorous test ofthe multimodal knowledge of vision-and-language models.

Do Vision-and-Language Transformers Learn Grounded Predicate-Noun Dependencies?