Model Agnostic Answer Reranking System for Adversarial Question Answering

TL;DR

This paper introduces a simple, model-agnostic answer reranking method for question answering systems that improves robustness against adversarial examples without retraining the models.

Contribution

The proposed approach is universally applicable to any QA model and outperforms existing defense techniques against adversarial attacks.

Findings

Outperforms state-of-the-art adversarial defenses

Does not require retraining of QA models

Effective across different QA architectures

Abstract

While numerous methods have been proposed as defenses against adversarial examples in question answering (QA), these techniques are often model specific, require retraining of the model, and give only marginal improvements in performance over vanilla models. In this work, we present a simple model-agnostic approach to this problem that can be applied directly to any QA model without any retraining. Our method employs an explicit answer candidate reranking mechanism that scores candidate answers on the basis of their content overlap with the question before making the final prediction. Combined with a strong base QAmodel, our method outperforms state-of-the-art defense techniques, calling into question how well these techniques are actually doing and strong these adversarial testbeds are.