# Uncertainty Calibration for Ensemble-Based Debiasing Methods

11/07/2021

∙

by [Ruibin Xiong, et al.](/content/profile/ruibin-xiong/index.html)

∙

0

Ensemble-based debiasing methods have been shown effective in mitigating the reliance of [classifiers](/content/machine-learning-glossary-and-terms/classifier/index.html) on specific dataset bias, by exploiting the output of a bias-only model to adjust the learning target. In this paper, we focus on the bias-only model in these ensemble-based methods, which plays an important role but has not gained much attention in the existing literature. Theoretically, we prove that the debiasing performance can be damaged by inaccurate uncertainty [estimations](/content/machine-learning-glossary-and-terms/estimator/index.html) of the bias-only model. Empirically, we show that existing bias-only models fall short in producing accurate uncertainty estimations. Motivated by these findings, we propose to conduct calibration on the bias-only model, thus achieving a three-stage ensemble-based debiasing framework, including bias modeling, model calibrating, and debiasing. Experimental results on NLI and fact verification tasks show that our proposed three-stage debiasing framework consistently outperforms the traditional two-stage one in out-of-distribution accuracy.

[READ FULL TEXT](http://arxiv.org/pdf/2111.04104v1)