Matches in SemOpenAlex for { <https://semopenalex.org/work/W4287068685> ?p ?o ?g. }
Showing items 1 to 87 of
87
with 100 items per page.
- W4287068685 abstract "Robustness of huge Transformer-based models for natural language processing is an important issue due to their capabilities and wide adoption. One way to understand and improve robustness of these models is an exploration of an adversarial attack scenario: check if a small perturbation of an input can fool a model. Due to the discrete nature of textual data, gradient-based adversarial methods, widely used in computer vision, are not applicable per~se. The standard strategy to overcome this issue is to develop token-level transformations, which do not take the whole sentence into account. In this paper, we propose a new black-box sentence-level attack. Our method fine-tunes a pre-trained language model to generate adversarial examples. A proposed differentiable loss function depends on a substitute classifier score and an approximate edit distance computed via a deep learning model. We show that the proposed attack outperforms competitors on a diverse set of NLP problems for both computed metrics and human evaluation. Moreover, due to the usage of the fine-tuned language model, the generated adversarial examples are hard to detect, thus current models are not robust. Hence, it is difficult to defend from the proposed attack, which is not the case for other attacks." @default.
- W4287068685 created "2022-07-25" @default.
- W4287068685 creator A5011784224 @default.
- W4287068685 creator A5011905327 @default.
- W4287068685 creator A5032790038 @default.
- W4287068685 creator A5047409045 @default.
- W4287068685 creator A5056308251 @default.
- W4287068685 creator A5066231807 @default.
- W4287068685 creator A5071539494 @default.
- W4287068685 creator A5088950452 @default.
- W4287068685 date "2021-07-23" @default.
- W4287068685 modified "2023-09-27" @default.
- W4287068685 title "A Differentiable Language Model Adversarial Attack on Text Classifiers" @default.
- W4287068685 doi "https://doi.org/10.48550/arxiv.2107.11275" @default.
- W4287068685 hasPublicationYear "2021" @default.
- W4287068685 type Work @default.
- W4287068685 citedByCount "0" @default.
- W4287068685 crossrefType "posted-content" @default.
- W4287068685 hasAuthorship W4287068685A5011784224 @default.
- W4287068685 hasAuthorship W4287068685A5011905327 @default.
- W4287068685 hasAuthorship W4287068685A5032790038 @default.
- W4287068685 hasAuthorship W4287068685A5047409045 @default.
- W4287068685 hasAuthorship W4287068685A5056308251 @default.
- W4287068685 hasAuthorship W4287068685A5066231807 @default.
- W4287068685 hasAuthorship W4287068685A5071539494 @default.
- W4287068685 hasAuthorship W4287068685A5088950452 @default.
- W4287068685 hasBestOaLocation W42870686851 @default.
- W4287068685 hasConcept C104317684 @default.
- W4287068685 hasConcept C119857082 @default.
- W4287068685 hasConcept C121332964 @default.
- W4287068685 hasConcept C134306372 @default.
- W4287068685 hasConcept C137293760 @default.
- W4287068685 hasConcept C154945302 @default.
- W4287068685 hasConcept C165801399 @default.
- W4287068685 hasConcept C185592680 @default.
- W4287068685 hasConcept C202615002 @default.
- W4287068685 hasConcept C204321447 @default.
- W4287068685 hasConcept C2777530160 @default.
- W4287068685 hasConcept C33923547 @default.
- W4287068685 hasConcept C37736160 @default.
- W4287068685 hasConcept C38652104 @default.
- W4287068685 hasConcept C41008148 @default.
- W4287068685 hasConcept C48145219 @default.
- W4287068685 hasConcept C55493867 @default.
- W4287068685 hasConcept C62520636 @default.
- W4287068685 hasConcept C63479239 @default.
- W4287068685 hasConcept C65856478 @default.
- W4287068685 hasConcept C66322947 @default.
- W4287068685 hasConcept C95623464 @default.
- W4287068685 hasConceptScore W4287068685C104317684 @default.
- W4287068685 hasConceptScore W4287068685C119857082 @default.
- W4287068685 hasConceptScore W4287068685C121332964 @default.
- W4287068685 hasConceptScore W4287068685C134306372 @default.
- W4287068685 hasConceptScore W4287068685C137293760 @default.
- W4287068685 hasConceptScore W4287068685C154945302 @default.
- W4287068685 hasConceptScore W4287068685C165801399 @default.
- W4287068685 hasConceptScore W4287068685C185592680 @default.
- W4287068685 hasConceptScore W4287068685C202615002 @default.
- W4287068685 hasConceptScore W4287068685C204321447 @default.
- W4287068685 hasConceptScore W4287068685C2777530160 @default.
- W4287068685 hasConceptScore W4287068685C33923547 @default.
- W4287068685 hasConceptScore W4287068685C37736160 @default.
- W4287068685 hasConceptScore W4287068685C38652104 @default.
- W4287068685 hasConceptScore W4287068685C41008148 @default.
- W4287068685 hasConceptScore W4287068685C48145219 @default.
- W4287068685 hasConceptScore W4287068685C55493867 @default.
- W4287068685 hasConceptScore W4287068685C62520636 @default.
- W4287068685 hasConceptScore W4287068685C63479239 @default.
- W4287068685 hasConceptScore W4287068685C65856478 @default.
- W4287068685 hasConceptScore W4287068685C66322947 @default.
- W4287068685 hasConceptScore W4287068685C95623464 @default.
- W4287068685 hasLocation W42870686851 @default.
- W4287068685 hasOpenAccess W4287068685 @default.
- W4287068685 hasPrimaryLocation W42870686851 @default.
- W4287068685 hasRelatedWork W3023124213 @default.
- W4287068685 hasRelatedWork W3036303027 @default.
- W4287068685 hasRelatedWork W3107474891 @default.
- W4287068685 hasRelatedWork W3205128835 @default.
- W4287068685 hasRelatedWork W4286899967 @default.
- W4287068685 hasRelatedWork W4287626481 @default.
- W4287068685 hasRelatedWork W4294962858 @default.
- W4287068685 hasRelatedWork W4307415490 @default.
- W4287068685 hasRelatedWork W4312741812 @default.
- W4287068685 hasRelatedWork W4318751489 @default.
- W4287068685 isParatext "false" @default.
- W4287068685 isRetracted "false" @default.
- W4287068685 workType "article" @default.