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

Test-time adaptation (TTA) aims to adapt a pre-trained model to the targetdomain in a batch-by-batch manner during inference. While label distributionsoften exhibit imbalances in real-world scenarios, most previous TTA approachestypically assume that both source and target domain datasets have balancedlabel distribution. Due to the fact that certain classes appear more frequentlyin certain domains (e.g., buildings in cities, trees in forests), it is naturalthat the label distribution shifts as the domain changes. However, we discoverthat the majority of existing TTA methods fail to address the coexistence ofcovariate and label shifts. To tackle this challenge, we propose a novel labelshift adapter that can be incorporated into existing TTA approaches to dealwith label shifts during the TTA process effectively. Specifically, we estimatethe label distribution of the target domain to feed it into the label shiftadapter. Subsequently, the label shift adapter produces optimal parameters forthe target label distribution. By predicting only the parameters for a part ofthe pre-trained source model, our approach is computationally efficient and canbe easily applied, regardless of the model architectures. Through extensiveexperiments, we demonstrate that integrating our strategy with TTA approachesleads to substantial performance improvements under the joint presence of labeland covariate shifts.

Label Shift Adapter for Test-Time Adaptation under Covariate and Label Shifts