EFSG: Evolutionary Fooling Sentences Generator

TL;DR

EFSG is an evolutionary algorithm that generates adversarial sentences to identify vulnerabilities in large language models like BERT and RoBERTa, and improves robustness through adversarial training.

Contribution

We introduce EFSG, a novel evolutionary approach for generating adversarial sentences applicable across models and tasks, revealing weaknesses and enhancing model robustness.

Findings

EFSG successfully fools BERT and RoBERTa on CoLA and MRPC tasks.

Adversarial training with EFSG data improves model robustness without accuracy loss.

State-of-the-art models have identifiable weak spots exposed by EFSG.

Abstract

Large pre-trained language representation models (LMs) have recently collected a huge number of successes in many NLP tasks. In 2018 BERT, and later its successors (e.g. RoBERTa), obtained state-of-the-art results in classical benchmark tasks, such as GLUE benchmark. After that, works about adversarial attacks have been published to test their generalization proprieties and robustness. In this work, we design Evolutionary Fooling Sentences Generator (EFSG), a model- and task-agnostic adversarial attack algorithm built using an evolutionary approach to generate false-positive sentences for binary classification tasks. We successfully apply EFSG to CoLA and MRPC tasks, on BERT and RoBERTa, comparing performances. Results prove the presence of weak spots in state-of-the-art LMs. We finally test adversarial training as a data augmentation defence approach against EFSG, obtaining stronger improved models with no loss of accuracy when tested on the original datasets.