RobustSentEmbed: Robust Sentence Embeddings Using Adversarial Self-Supervised Contrastive Learning

TL;DR

RobustSentEmbed is a novel self-supervised framework that enhances the robustness and generalization of sentence embeddings against adversarial attacks, significantly reducing attack success rates and improving performance on multiple NLP tasks.

Contribution

It introduces a new adversarial self-supervised contrastive learning method for robust sentence embeddings, addressing robustness issues in PLM-based representations.

Findings

Reduces BERTAttack success rate from 75.51% to 38.81%.

Improves semantic textual similarity by 1.59%.

Enhances transfer task performance by 0.23%.

Abstract

Pre-trained language models (PLMs) have consistently demonstrated outstanding performance across a diverse spectrum of natural language processing tasks. Nevertheless, despite their success with unseen data, current PLM-based representations often exhibit poor robustness in adversarial settings. In this paper, we introduce RobustSentEmbed, a self-supervised sentence embedding framework designed to improve both generalization and robustness in diverse text representation tasks and against a diverse set of adversarial attacks. Through the generation of high-risk adversarial perturbations and their utilization in a novel objective function, RobustSentEmbed adeptly learns high-quality and robust sentence embeddings. Our experiments confirm the superiority of RobustSentEmbed over state-of-the-art representations. Specifically, Our framework achieves a significant reduction in the success rate of various adversarial attacks, notably reducing the BERTAttack success rate by almost half (from 75.51\% to 38.81\%). The framework also yields improvements of 1.59\% and 0.23\% in semantic textual similarity tasks and various transfer tasks, respectively.