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

The heightened emphasis on the regulation of deep generative models,propelled by escalating concerns pertaining to privacy and compliance withregulatory frameworks, underscores the imperative need for precise controlmechanisms over these models. This urgency is particularly underscored byinstances in which generative models generate outputs that encompassobjectionable, offensive, or potentially injurious content. In response,machine unlearning has emerged to selectively forget specific knowledge orremove the influence of undesirable data subsets from pre-trained models.However, modern machine unlearning approaches typically assume access to modelparameters and architectural details during unlearning, which is not alwaysfeasible. In multitude of downstream tasks, these models function as black-boxsystems, with inaccessible pre-trained parameters, architectures, and trainingdata. In such scenarios, the possibility of filtering undesired outputs becomesa practical alternative. The primary goal of this study is twofold: first, toelucidate the relationship between filtering and unlearning processes, andsecond, to formulate a methodology aimed at mitigating the display ofundesirable outputs generated from models characterized as black-box systems.Theoretical analysis in this study demonstrates that, in the context ofblack-box models, filtering can be seen as a form of weak unlearning. Ourproposed \textbf{\textit{Feature Aware Similarity Thresholding(FAST)}} methodeffectively suppresses undesired outputs by systematically encoding therepresentation of unwanted features in the latent space.

FAST: Feature Aware Similarity Thresholding for Weak Unlearning in Black-Box Generative Models