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 much success in natural language processing (NLP), pre-trainedlanguage models typically lead to a high computational cost during inference.Multi-exit is a mainstream approach to address this issue by making a trade-offbetween efficiency and accuracy, where the saving of computation comes from anearly exit. However, whether such saving from early-exiting is robust remainsunknown. Motivated by this, we first show that directly adapting existingadversarial attack approaches targeting model accuracy cannot significantlyreduce inference efficiency. To this end, we propose a simple yet effectiveattacking framework, SAME, a novel slowdown attack framework on multi-exitmodels, which is specially tailored to reduce the efficiency of the multi-exitmodels. By leveraging the multi-exit models' design characteristics, we utilizeall internal predictions to guide the adversarial sample generation instead ofmerely considering the final prediction. Experiments on the GLUE benchmark showthat SAME can effectively diminish the efficiency gain of various multi-exitmodels by 80% on average, convincingly validating its effectiveness andgeneralization ability.

Dynamic Transformers Provide a False Sense of Efficiency