Table 5 xFiTRNNs’ adversarial robustness on financial phrasebank.

From: A hybrid self attentive linearized phrase structured transformer based RNN for financial sentence analysis with sentence level explainability

Attack type

Success rate

F1 drop

Glyph substitution

(“growth”\(\rightarrow\)“groWth”)

12.71%

3.23%

Financial negation insertion

(“strong results”\(\rightarrow\)“not strong results”)

18.44%

5.18%

SEC filing obfuscation

(Adding legalese)

9.36%

2.15%