A Unified Evaluation of Textual Backdoor Learning: Frameworks and Benchmarks

1 code implementation • 17 Jun 2022 • Ganqu Cui, Lifan Yuan, Bingxiang He, Yangyi Chen, Zhiyuan Liu, Maosong Sun

However, we highlight two issues in previous backdoor learning evaluations: (1) The differences between real-world scenarios (e. g. releasing poisoned datasets or models) are neglected, and we argue that each scenario has its own constraints and concerns, thus requires specific evaluation protocols; (2) The evaluation metrics only consider whether the attacks could flip the models' predictions on poisoned samples and retain performances on benign samples, but ignore that poisoned samples should also be stealthy and semantic-preserving.

text similarity