Hybrid entropy-uncertainty-geometric defence improves clean accuracy by up to 43% and adversarial robustness by up to 65% on NLU and security benchmarks.
Defending Pre-trained Language Models from Adversarial Word Substitution Without Performance Sacrifice
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Hybrid Adversarial Defence for Natural Language Understanding Tasks
Hybrid entropy-uncertainty-geometric defence improves clean accuracy by up to 43% and adversarial robustness by up to 65% on NLU and security benchmarks.