Exploring the Universal Vulnerability of Prompt-based Learning Paradigm

Anonymous

Paper Link: https://openreview.net/forum?id=OFzvrBameCo

Paper Type: Long paper (up to eight pages of content + unlimited references and appendices)

Abstract: Prompt-based learning paradigm bridges the gap between pre-training and fine-tuning, and works effectively under the few-shot setting. However, we find that this learning paradigm inherits the vulnerability from the pre-training stage, where model predictions can be misled by inserting certain triggers into the text. In this paper, we explore this universal vulnerability by either injecting backdoor triggers or searching for adversarial triggers on pre-trained language models using only plain text. In both scenarios, we demonstrate that our triggers can totally control or severely decrease the performance of prompt-based models fine-tuned on arbitrary downstream tasks, reflecting the universal vulnerability of the prompt-based learning paradigm. Further experiments show that adversarial triggers have good transferability among language models. We also find conventional fine-tuning models are not vulnerable to adversarial triggers constructed from pre-trained language models. We conclude by proposing a potential solution to mitigate our attack methods. Code and data are publicly available.

Presentation Mode: This paper will be presented in person in Seattle

Virtual Presentation Timezone: UTC-4

Copyright Consent Signature (type Name Or NA If Not Transferrable): Lei Xu

Copyright Consent Name And Address: MIT LIDS; 77 Massachusetts Ave, Cambridge, MA, USA