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

Recent work has focused on compressing pre-trained language models (PLMs)like BERT where the major focus has been to improve the in-distributionperformance for downstream tasks. However, very few of these studies haveanalyzed the impact of compression on the generalizability and robustness ofcompressed models for out-of-distribution (OOD) data. Towards this end, westudy two popular model compression techniques including knowledge distillationand pruning and show that the compressed models are significantly less robustthan their PLM counterparts on OOD test sets although they obtain similarperformance on in-distribution development sets for a task. Further analysisindicates that the compressed models overfit on the shortcut samples andgeneralize poorly on the hard ones. We further leverage this observation todevelop a regularization strategy for robust model compression based on sampleuncertainty. Experimental results on several natural language understandingtasks demonstrate that our bias mitigation framework improves the OODgeneralization of the compressed models, while not sacrificing thein-distribution task performance.

Robustness Challenges in Model Distillation and Pruning for Natural Language Understanding