longhorns at DADC 2022: How many linguists does it take to fool a Question Answering model? A systematic approach to adversarial attacks

2022/06/29 by Venelin Kovatchev, Kovatchev, Venelin, Trina Chatterjee +21 · 1 citation
Computer Science · Health Professions · #Topic Modeling #Natural Language Processing Techniques #Interpreting and Communication in Healthcare

paper · pdf · doi:10.48550/arxiv.2206.14729

Abstract

Developing methods to adversarially challenge NLP systems is a promising avenue for improving both model performance and interpretability. Here, we describe the approach of the team "longhorns" on Task 1 of the The First Workshop on Dynamic Adversarial Data Collection (DADC), which asked teams to manually fool a model on an Extractive Question Answering task. Our team finished first, with a model error rate of 62%. We advocate for a systematic, linguistically informed approach to formulating adversarial questions, and we describe the results of our pilot experiments, as well as our official submission.

Cited by

• Revisiting CroPA: A Reproducibility Study and Enhancements for Cross-Prompt Adversarial Transferability in Vision-Language Models

Related

• How the Advent of Ubiquitous Large Language Models both Stymie and Turbocharge Dynamic Adversarial Question Generation
• Deceiving Question-Answering Models: A Hybrid Word-Level Adversarial Approach
• Trick Me If You Can: Human-in-the-loop Generation of Adversarial Examples for Question Answering
• On the Efficacy of Adversarial Data Collection for Question Answering:\n Results from a Large-Scale Randomized Study
• Adversarial Attacks and Defense for Conversation Entailment Task