Matches in SemOpenAlex for { <https://semopenalex.org/work/W3035164976> ?p ?o ?g. }
Showing items 1 to 84 of
84
with 100 items per page.
- W3035164976 abstract "State-of-the-art NLP models can often be fooled by human-unaware transformations such as synonymous word substitution. For security reasons, it is of critical importance to develop models with certified robustness that can provably guarantee that the prediction is can not be altered by any possible synonymous word substitution. In this work, we propose a certified robust method based on a new randomized smoothing technique, which constructs a stochastic ensemble by applying random word substitutions on the input sentences, and leverage the statistical properties of the ensemble to provably certify the robustness. Our method is simple and structure-free in that it only requires the black-box queries of the model outputs, and hence can be applied to any pre-trained models (such as BERT) and any types of models (world-level or subword-level). Our method significantly outperforms recent state-of-the-art methods for certified robustness on both IMDB and Amazon text classification tasks. To the best of our knowledge, we are the first work to achieve certified robustness on large systems such as BERT with practically meaningful certified accuracy." @default.
- W3035164976 created "2020-06-19" @default.
- W3035164976 creator A5011339261 @default.
- W3035164976 creator A5040717619 @default.
- W3035164976 creator A5057055806 @default.
- W3035164976 date "2020-01-01" @default.
- W3035164976 modified "2023-10-16" @default.
- W3035164976 title "SAFER: A Structure-free Approach for Certified Robustness to Adversarial Word Substitutions" @default.
- W3035164976 cites W1832693441 @default.
- W3035164976 cites W2061873838 @default.
- W3035164976 cites W2113459411 @default.
- W3035164976 cites W2250539671 @default.
- W3035164976 cites W2493916176 @default.
- W3035164976 cites W2799007037 @default.
- W3035164976 cites W2799194071 @default.
- W3035164976 cites W2803392236 @default.
- W3035164976 cites W2911634294 @default.
- W3035164976 cites W2962718684 @default.
- W3035164976 cites W2963012544 @default.
- W3035164976 cites W2963341956 @default.
- W3035164976 cites W2964073004 @default.
- W3035164976 cites W2964232431 @default.
- W3035164976 cites W2965595599 @default.
- W3035164976 cites W2969442125 @default.
- W3035164976 cites W2970078867 @default.
- W3035164976 cites W2970449623 @default.
- W3035164976 cites W2971109239 @default.
- W3035164976 cites W2990740676 @default.
- W3035164976 cites W3008683143 @default.
- W3035164976 doi "https://doi.org/10.18653/v1/2020.acl-main.317" @default.
- W3035164976 hasPublicationYear "2020" @default.
- W3035164976 type Work @default.
- W3035164976 sameAs 3035164976 @default.
- W3035164976 citedByCount "35" @default.
- W3035164976 countsByYear W30351649762019 @default.
- W3035164976 countsByYear W30351649762020 @default.
- W3035164976 countsByYear W30351649762021 @default.
- W3035164976 countsByYear W30351649762022 @default.
- W3035164976 countsByYear W30351649762023 @default.
- W3035164976 crossrefType "proceedings-article" @default.
- W3035164976 hasAuthorship W3035164976A5011339261 @default.
- W3035164976 hasAuthorship W3035164976A5040717619 @default.
- W3035164976 hasAuthorship W3035164976A5057055806 @default.
- W3035164976 hasBestOaLocation W30351649761 @default.
- W3035164976 hasConcept C104317684 @default.
- W3035164976 hasConcept C119857082 @default.
- W3035164976 hasConcept C153083717 @default.
- W3035164976 hasConcept C154945302 @default.
- W3035164976 hasConcept C185592680 @default.
- W3035164976 hasConcept C204321447 @default.
- W3035164976 hasConcept C31972630 @default.
- W3035164976 hasConcept C3770464 @default.
- W3035164976 hasConcept C41008148 @default.
- W3035164976 hasConcept C55493867 @default.
- W3035164976 hasConcept C63479239 @default.
- W3035164976 hasConceptScore W3035164976C104317684 @default.
- W3035164976 hasConceptScore W3035164976C119857082 @default.
- W3035164976 hasConceptScore W3035164976C153083717 @default.
- W3035164976 hasConceptScore W3035164976C154945302 @default.
- W3035164976 hasConceptScore W3035164976C185592680 @default.
- W3035164976 hasConceptScore W3035164976C204321447 @default.
- W3035164976 hasConceptScore W3035164976C31972630 @default.
- W3035164976 hasConceptScore W3035164976C3770464 @default.
- W3035164976 hasConceptScore W3035164976C41008148 @default.
- W3035164976 hasConceptScore W3035164976C55493867 @default.
- W3035164976 hasConceptScore W3035164976C63479239 @default.
- W3035164976 hasLocation W30351649761 @default.
- W3035164976 hasLocation W30351649762 @default.
- W3035164976 hasOpenAccess W3035164976 @default.
- W3035164976 hasPrimaryLocation W30351649761 @default.
- W3035164976 hasRelatedWork W1978572805 @default.
- W3035164976 hasRelatedWork W1987225439 @default.
- W3035164976 hasRelatedWork W1989941125 @default.
- W3035164976 hasRelatedWork W1997992934 @default.
- W3035164976 hasRelatedWork W2019977573 @default.
- W3035164976 hasRelatedWork W2125114371 @default.
- W3035164976 hasRelatedWork W2149980199 @default.
- W3035164976 hasRelatedWork W2383807498 @default.
- W3035164976 hasRelatedWork W2391139946 @default.
- W3035164976 hasRelatedWork W3125766170 @default.
- W3035164976 isParatext "false" @default.
- W3035164976 isRetracted "false" @default.
- W3035164976 magId "3035164976" @default.
- W3035164976 workType "article" @default.