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

Beyond the success story of adversarial training (AT) in the recent textdomain on top of pre-trained language models (PLMs), our empirical studyshowcases the inconsistent gains from AT on some tasks, e.g. commonsensereasoning, named entity recognition. This paper investigates AT from theperspective of the contextualized language representation outputted by PLMencoders. We find the current AT attacks lean to generate sub-optimaladversarial examples that can fool the decoder part but have a minor effect onthe encoder. However, we find it necessary to effectively deviate the latterone to allow AT to gain. Based on the observation, we propose simple yeteffective \textit{Contextualized representation-Adversarial Training} (CreAT),in which the attack is explicitly optimized to deviate the contextualizedrepresentation of the encoder. It allows a global optimization of adversarialexamples that can fool the entire model. We also find CreAT gives rise to abetter direction to optimize the adversarial examples, to let them lesssensitive to hyperparameters. Compared to AT, CreAT produces consistentperformance gains on a wider range of tasks and is proven to be more effectivefor language pre-training where only the encoder part is kept for downstreamtasks. We achieve the new state-of-the-art performances on a series ofchallenging benchmarks, e.g. AdvGLUE (59.1 $ \rightarrow $ 61.1), HellaSWAG(93.0 $ \rightarrow $ 94.9), ANLI (68.1 $ \rightarrow $ 69.3).

Toward Adversarial Training on Contextualized Language Representation