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

Despite advances in Visual Question Answering (VQA), the ability of models toassess their own correctness remains underexplored. Recent work has shown thatVQA models, out-of-the-box, can have difficulties abstaining from answeringwhen they are wrong. The option to abstain, also called Selective Prediction,is highly relevant when deploying systems to users who must trust the system'soutput (e.g., VQA assistants for users with visual impairments). For suchscenarios, abstention can be especially important as users may provideout-of-distribution (OOD) or adversarial inputs that make incorrect answersmore likely. In this work, we explore Selective VQA in both in-distribution(ID) and OOD scenarios, where models are presented with mixtures of ID and OODdata. The goal is to maximize the number of questions answered while minimizingthe risk of error on those questions. We propose a simple yet effectiveLearning from Your Peers (LYP) approach for training multimodal selectionfunctions for making abstention decisions. Our approach uses predictions frommodels trained on distinct subsets of the training data as targets foroptimizing a Selective VQA model. It does not require additional manual labelsor held-out data and provides a signal for identifying examples that areeasy/difficult to generalize to. In our extensive evaluations, we show thisbenefits a number of models across different architectures and scales. Overall,for ID, we reach 32.92% in the selective prediction metric coverage at 1% riskof error (C@1%) which doubles the previous best coverage of 15.79% on thistask. For mixed ID/OOD, using models' softmax confidences for abstentiondecisions performs very poorly, answering <5% of questions at 1% risk of erroreven when faced with only 10% OOD examples, but a learned selection functionwith LYP can increase that to 25.38% C@1%.

Improving Selective Visual Question Answering by Learning from Your Peers