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

Attention maps, a popular heatmap-based explanation method for VisualQuestion Answering (VQA), are supposed to help users understand the model byhighlighting portions of the image/question used by the model to infer answers.However, we see that users are often misled by current attention mapvisualizations that point to relevant regions despite the model producing anincorrect answer. Hence, we propose Error Maps that clarify the error byhighlighting image regions where the model is prone to err. Error maps canindicate when a correctly attended region may be processed incorrectly leadingto an incorrect answer, and hence, improve users' understanding of those cases.To evaluate our new explanations, we further introduce a metric that simulatesusers' interpretation of explanations to evaluate their potential helpfulnessto understand model correctness. We finally conduct user studies to see thatour new explanations help users understand model correctness better thanbaselines by an expected 30\% and that our proxy helpfulness metrics correlatestrongly ($\rho>0.97$) with how well users can predict model correctness.

Generating and Evaluating Explanations of Attended and Error-Inducing Input Regions for VQA Models