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

Despite the subjective nature of many NLP tasks, most NLU evaluations havefocused on using the majority label with presumably high agreement as theground truth. Less attention has been paid to the distribution of humanopinions. We collect ChaosNLI, a dataset with a total of 464,500 annotations tostudy Collective HumAn OpinionS in oft-used NLI evaluation sets. This datasetis created by collecting 100 annotations per example for 3,113 examples in SNLIand MNLI and 1,532 examples in Abductive-NLI. Analysis reveals that: (1) highhuman disagreement exists in a noticeable amount of examples in these datasets;(2) the state-of-the-art models lack the ability to recover the distributionover human labels; (3) models achieve near-perfect accuracy on the subset ofdata with a high level of human agreement, whereas they can barely beat arandom guess on the data with low levels of human agreement, which compose mostof the common errors made by state-of-the-art models on the evaluation sets.This questions the validity of improving model performance on old metrics forthe low-agreement part of evaluation datasets. Hence, we argue for a detailedexamination of human agreement in future data collection efforts, andevaluating model outputs against the distribution over collective humanopinions. The ChaosNLI dataset and experimental scripts are available athttps://github.com/easonnie/ChaosNLI

What Can We Learn from Collective Human Opinions on Natural Language Inference Data?