VoteTRANS: Detecting Adversarial Text without Training by Voting on Hard Labels of Transformations.
Hoang-Quoc Nguyen-SonSeira HidanoKazuhide FukushimaShinsaku KiyomotoIsao EchizenPublished in: ACL (Findings) (2023)
Keyphrases
- training examples
- information retrieval
- training set
- labelled data
- database
- active learning
- annotation effort
- keywords
- training algorithm
- data sets
- ground truth labels
- voting protocols
- textual data
- training process
- text retrieval
- test set
- web documents
- training samples
- pairwise
- free text
- text data
- text documents
- training corpus
- labeling effort
- multi agent