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

Fine-tuning pre-trained language models (PTLMs), such as BERT and its bettervariant RoBERTa, has been a common practice for advancing performance innatural language understanding (NLU) tasks. Recent advance in representationlearning shows that isotropic (i.e., unit-variance and uncorrelated) embeddingscan significantly improve performance on downstream tasks with fasterconvergence and better generalization. The isotropy of the pre-trainedembeddings in PTLMs, however, is relatively under-explored. In this paper, weanalyze the isotropy of the pre-trained [CLS] embeddings of PTLMs withstraightforward visualization, and point out two major issues: high variance intheir standard deviation, and high correlation between different dimensions. Wealso propose a new network regularization method, isotropic batch normalization(IsoBN) to address the issues, towards learning more isotropic representationsin fine-tuning by dynamically penalizing dominating principal components. Thissimple yet effective fine-tuning method yields about 1.0 absolute increment onthe average of seven NLU tasks.

IsoBN: Fine-Tuning BERT with Isotropic Batch Normalization