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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Neural networks often learn spurious correlations when exposed to biasedtraining data, leading to poor performance on out-of-distribution data. Abiased dataset can be divided, according to biased features, into bias-alignedsamples (i.e., with biased features) and bias-conflicting samples (i.e.,without biased features). Recent debiasing works typically assume that no biaslabel is available during the training phase, as obtaining such information ischallenging and labor-intensive. Following this unsupervised assumption,existing methods usually train two models: a biased model specialized to learnbiased features and a target model that uses information from the biased modelfor debiasing. This paper first presents experimental analyses revealing thatthe existing biased models overfit to bias-conflicting samples in the trainingdata, which negatively impacts the debiasing performance of the target models.To address this issue, we propose a straightforward and effective method calledEchoes, which trains a biased model and a target model with a differentstrategy. We construct an "echo chamber" environment by reducing the weights ofsamples which are misclassified by the biased model, to ensure the biased modelfully learns the biased features without overfitting to the bias-conflictingsamples. The biased model then assigns lower weights on the bias-conflictingsamples. Subsequently, we use the inverse of the sample weights of the biasedmodel for training the target model. Experiments show that our approachachieves superior debiasing results compared to the existing baselines on bothsynthetic and real-world datasets. Our code is available athttps://github.com/isruihu/Echoes.

Echoes: Unsupervised Debiasing via Pseudo-bias Labeling in an Echo Chamber