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

Cross-task generalization is a significant outcome that defines mastery innatural language understanding. Humans show a remarkable aptitude for this, andcan solve many different types of tasks, given definitions in the form oftextual instructions and a small set of examples. Recent work with pre-trainedlanguage models mimics this learning style: users can define and exemplify atask for the model to attempt as a series of natural language prompts orinstructions. While prompting approaches have led to higher cross-taskgeneralization compared to traditional supervised learning, analyzing 'bias' inthe task instructions given to the model is a difficult problem, and has thusbeen relatively unexplored. For instance, are we truly modeling a task, or arewe modeling a user's instructions? To help investigate this, we develop LINGO,a novel visual analytics interface that supports an effective, task-drivenworkflow to (1) help identify bias in natural language task instructions, (2)alter (or create) task instructions to reduce bias, and (3) evaluatepre-trained model performance on debiased task instructions. To robustlyevaluate LINGO, we conduct a user study with both novice and expert instructioncreators, over a dataset of 1,616 linguistic tasks and their natural languageinstructions, spanning 55 different languages. For both user groups, LINGOpromotes the creation of more difficult tasks for pre-trained models, thatcontain higher linguistic diversity and lower instruction bias. We additionallydiscuss how the insights learned in developing and evaluating LINGO can aid inthe design of future dashboards that aim to minimize the effort involved inprompt creation across multiple domains.

LINGO : Visually Debiasing Natural Language Instructions to Support Task Diversity