Learning to Discriminate Adversarial Examples by Sensitivity Inconsistency in IoHT Systems

Table 8

Performance of detector on different datasets.

Dataset	BERT + Deepwordbug		LSTM + TextFooler
	Recall (%)	F1-score (%)	Recall (%)	F1-score (%)
IMDB	94.2	94.8	97.2	96.4
AG’s news	85.0	85.6	96.2	91.2
Yelp	90.1	92	95.5	93.4
SST-2	81.2	84.3	87.6	89.1