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

As AI models rapidly evolve, they are frequently released to openrepositories, such as HuggingFace. It is essential to perform quality assurancevalidation on these models before integrating them into the productiondevelopment lifecycle. In addition to evaluating efficiency in terms ofbalanced accuracy and computing costs, adversarial attacks are potentialthreats to the robustness and explainability of AI models. Meanwhile, XAIapplies algorithms that approximate inputs to outputs post-hoc to identify thecontributing features. Adversarial perturbations may also degrade the utilityof XAI explanations that require further investigation. In this paper, wepresent an integrated process designed for downstream evaluation tasks,including validating AI model accuracy, evaluating robustness with benchmarkperturbations, comparing explanation utility, and assessing overhead. Wedemonstrate an evaluation scenario involving six computer vision models, whichinclude CNN-based, Transformer-based, and hybrid architectures, three types ofperturbations, and five XAI methods, resulting in ninety unique combinations.The process reveals the explanation utility among the XAI methods in terms ofthe identified key areas responding to the adversarial perturbation. Theprocess produces aggregated results that illustrate multiple attributes of eachAI model.

Analyzing the Quality Attributes of AI Vision Models in Open Repositories Under Adversarial Attacks