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

Multilingual transformer-based models demonstrate remarkable zero andfew-shot transfer across languages by learning and reusing language-agnosticfeatures. However, as a fixed-size model acquires more languages, itsperformance across all languages degrades, a phenomenon termed interference.Often attributed to limited model capacity, interference is commonly addressedby adding additional parameters despite evidence that transformer-based modelsare overparameterized. In this work, we show that it is possible to reduceinterference by instead identifying and pruning language-specific parameters.First, we use Shapley Values, a credit allocation metric from coalitional gametheory, to identify attention heads that introduce interference. Then, we showthat removing identified attention heads from a fixed model improvesperformance for a target language on both sentence classification andstructural prediction, seeing gains as large as 24.7\%. Finally, we provideinsights on language-agnostic and language-specific attention heads usingattention visualization.

Shapley Head Pruning: Identifying and Removing Interference in Multilingual Transformers