Abstract: Recent works have demonstrated that text classifiers are vulnerable to universal adversarial triggers (UATs), which are concatenated to any original text from the dataset to mislead text ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback