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 Tools annual

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 monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Teaching Visual Question Answering (VQA) models to refrain from answeringunanswerable questions is necessary for building a trustworthy AI system.Existing studies, though have explored various aspects of VQA but somewhatignored this particular attribute. This paper aims to bridge the research gapby contributing a comprehensive dataset, called UNK-VQA. The dataset isspecifically designed to address the challenge of questions that models do notknow. To this end, we first augment the existing data via deliberateperturbations on either the image or question. In specific, we carefully ensurethat the question-image semantics remain close to the original unperturbeddistribution. By this means, the identification of unanswerable questionsbecomes challenging, setting our dataset apart from others that involve mereimage replacement. We then extensively evaluate the zero- and few-shotperformance of several emerging multi-modal large models and discover theirsignificant limitations when applied to our dataset. Additionally, we alsopropose a straightforward method to tackle these unanswerable questions. Thisdataset, we believe, will serve as a valuable benchmark for enhancing theabstention capability of VQA models, thereby leading to increasedtrustworthiness of AI systems. We have made the\href{https://github.com/guoyang9/UNK-VQA}{dataset} available to facilitatefurther exploration in this area.

UNK-VQA: A Dataset and A Probe into Multi-modal Large Models' Abstention Ability