Think Twice: Measuring the Efficiency of Eliminating Prediction Shortcuts of Question Answering Models

TL;DR

This paper introduces a method to measure how much question answering models depend on spurious dataset biases, revealing that current debiasing techniques do not fully address shared biases across datasets, impacting model robustness evaluation.

Contribution

The authors propose a simple, scalable method to quantify models' reliance on known and new spurious features in QA, highlighting shared biases and limitations of existing debiasing approaches.

Findings

Debiasing methods reduce reliance on targeted spurious features.

OOD performance improvements are not solely due to bias mitigation.

Models trained on different datasets depend on similar bias features.

Abstract

While the Large Language Models (LLMs) dominate a majority of language understanding tasks, previous work shows that some of these results are supported by modelling spurious correlations of training datasets. Authors commonly assess model robustness by evaluating their models on out-of-distribution (OOD) datasets of the same task, but these datasets might share the bias of the training dataset. We propose a simple method for measuring a scale of models' reliance on any identified spurious feature and assess the robustness towards a large set of known and newly found prediction biases for various pre-trained models and debiasing methods in Question Answering (QA). We find that while existing debiasing methods can mitigate reliance on a chosen spurious feature, the OOD performance gains of these methods can not be explained by mitigated reliance on biased features, suggesting that biases are shared among different QA datasets. Finally, we evidence this to be the case by measuring that the performance of models trained on different QA datasets relies comparably on the same bias features. We hope these results will motivate future work to refine the reports of LMs' robustness to a level of adversarial samples addressing specific spurious features.