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

Despite significant success in Visual Question Answering (VQA), VQA modelshave been shown to be notoriously brittle to linguistic variations in thequestions. Due to deficiencies in models and datasets, today's models oftenrely on correlations rather than predictions that are causal w.r.t. data. Inthis paper, we propose a novel way to analyze and measure the robustness of thestate of the art models w.r.t semantic visual variations as well as proposeways to make models more robust against spurious correlations. Our methodperforms automated semantic image manipulations and tests for consistency inmodel predictions to quantify the model robustness as well as generatesynthetic data to counter these problems. We perform our analysis on threediverse, state of the art VQA models and diverse question types with aparticular focus on challenging counting questions. In addition, we show thatmodels can be made significantly more robust against inconsistent predictionsusing our edited data. Finally, we show that results also translate toreal-world error cases of state of the art models, which results in improvedoverall performance.

Towards Causal VQA: Revealing and Reducing Spurious Correlations by Invariant and Covariant Semantic Editing