A Jacobian-magnitude regularization called GBM improves empirical robustness of CNN/LSTM/S4 text classifiers to synonym-substitution attacks, but the claimed certified robustness is not delivered.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Bridging Robustness and Generalization Against Word Substitution Attacks in NLP via the Growth Bound Matrix Approach
A Jacobian-magnitude regularization called GBM improves empirical robustness of CNN/LSTM/S4 text classifiers to synonym-substitution attacks, but the claimed certified robustness is not delivered.