Adversarial GLUE: A Multi-Task Benchmark for Robustness Evaluation of Language Models

by   Boxin Wang, et al.
Zhejiang University
University of Illinois at Urbana-Champaign

Large-scale pre-trained language models have achieved tremendous success across a wide range of natural language understanding (NLU) tasks, even surpassing human performance. However, recent studies reveal that the robustness of these models can be challenged by carefully crafted textual adversarial examples. While several individual datasets have been proposed to evaluate model robustness, a principled and comprehensive benchmark is still missing. In this paper, we present Adversarial GLUE (AdvGLUE), a new multi-task benchmark to quantitatively and thoroughly explore and evaluate the vulnerabilities of modern large-scale language models under various types of adversarial attacks. In particular, we systematically apply 14 textual adversarial attack methods to GLUE tasks to construct AdvGLUE, which is further validated by humans for reliable annotations. Our findings are summarized as follows. (i) Most existing adversarial attack algorithms are prone to generating invalid or ambiguous adversarial examples, with around 90 either changing the original semantic meanings or misleading human annotators as well. Therefore, we perform a careful filtering process to curate a high-quality benchmark. (ii) All the language models and robust training methods we tested perform poorly on AdvGLUE, with scores lagging far behind the benign accuracy. We hope our work will motivate the development of new adversarial attacks that are more stealthy and semantic-preserving, as well as new robust language models against sophisticated adversarial attacks. AdvGLUE is available at


Lost In Translation: Generating Adversarial Examples Robust to Round-Trip Translation

Language Models today provide a high accuracy across a large number of d...

DecodingTrust: A Comprehensive Assessment of Trustworthiness in GPT Models

Generative Pre-trained Transformer (GPT) models have exhibited exciting ...

Semantic-Preserving Adversarial Code Comprehension

Based on the tremendous success of pre-trained language models (PrLMs) f...

Preserving Semantics in Textual Adversarial Attacks

Adversarial attacks in NLP challenge the way we look at language models....

A Prompting-based Approach for Adversarial Example Generation and Robustness Enhancement

Recent years have seen the wide application of NLP models in crucial are...

Defending Pre-trained Language Models from Adversarial Word Substitutions Without Performance Sacrifice

Pre-trained contextualized language models (PrLMs) have led to strong pe...

Task Ambiguity in Humans and Language Models

Language models have recently achieved strong performance across a wide ...

Please sign up or login with your details

Forgot password? Click here to reset