FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2972051251> ?p ?o ?g. }

• W2972051251 abstract "State-of-the-art models often make use of superficial patterns in the data that do not generalize well to out-of-domain or adversarial settings. For example, textual entailment models often learn that particular key words imply entailment, irrespective of context, and visual question answering models learn to predict prototypical answers, without considering evidence in the image. In this paper, we show that if we have prior knowledge of such biases, we can train a model to be more robust to domain shift. Our method has two stages: we (1) train a naive model that makes predictions exclusively based on dataset biases, and (2) train a robust model as part of an ensemble with the naive one in order to encourage it to focus on other patterns in the data that are more likely to generalize. Experiments on five datasets with out-of-domain test sets show significantly improved robustness in all settings, including a 12 point gain on a changing priors visual question answering dataset and a 9 point gain on an adversarial question answering test set." @default.
• W2972051251 created "2019-09-12" @default.
• W2972051251 creator A5020085226 @default.
• W2972051251 creator A5039205261 @default.
• W2972051251 creator A5067919401 @default.
• W2972051251 date "2019-09-09" @default.
• W2972051251 modified "2023-09-27" @default.
• W2972051251 title "Don't Take the Easy Way Out: Ensemble Based Methods for Avoiding Known Dataset Biases." @default.
• W2972051251 cites W1840435438 @default.
• W2972051251 cites W1933349210 @default.
• W2972051251 cites W2022166150 @default.
• W2972051251 cites W2096765155 @default.
• W2972051251 cites W2116064496 @default.
• W2972051251 cites W2167216307 @default.
• W2972051251 cites W2471094925 @default.
• W2972051251 cites W2547185913 @default.
• W2972051251 cites W2551396370 @default.
• W2972051251 cites W2745461083 @default.
• W2972051251 cites W2794325560 @default.
• W2972051251 cites W2794583223 @default.
• W2972051251 cites W2798665661 @default.
• W2972051251 cites W2888302696 @default.
• W2972051251 cites W2889787757 @default.
• W2972051251 cites W2898662126 @default.
• W2972051251 cites W2901716912 @default.
• W2972051251 cites W2901725278 @default.
• W2972051251 cites W2913222130 @default.
• W2972051251 cites W2942128719 @default.
• W2972051251 cites W2950470622 @default.
• W2972051251 cites W2950761309 @default.
• W2972051251 cites W2951286828 @default.
• W2972051251 cites W2953163841 @default.
• W2972051251 cites W2961158440 @default.
• W2972051251 cites W2962685807 @default.
• W2972051251 cites W2962718483 @default.
• W2972051251 cites W2962736243 @default.
• W2972051251 cites W2962772361 @default.
• W2972051251 cites W2962787423 @default.
• W2972051251 cites W2963053914 @default.
• W2972051251 cites W2963080779 @default.
• W2972051251 cites W2963096121 @default.
• W2972051251 cites W2963116854 @default.
• W2972051251 cites W2963140463 @default.
• W2972051251 cites W2963159690 @default.
• W2972051251 cites W2963339397 @default.
• W2972051251 cites W2963341956 @default.
• W2972051251 cites W2963349562 @default.
• W2972051251 cites W2963350032 @default.
• W2972051251 cites W2963360627 @default.
• W2972051251 cites W2963644680 @default.
• W2972051251 cites W2963661177 @default.
• W2972051251 cites W2963748441 @default.
• W2972051251 cites W2963783970 @default.
• W2972051251 cites W2963866616 @default.
• W2972051251 cites W2963890019 @default.
• W2972051251 cites W2963969878 @default.
• W2972051251 cites W2964121744 @default.
• W2972051251 cites W2964150944 @default.
• W2972051251 cites W2964153729 @default.
• W2972051251 cites W2964223283 @default.
• W2972051251 cites W2964301649 @default.
• W2972051251 cites W2964308564 @default.
• W2972051251 cites W2970730986 @default.
• W2972051251 cites W3016211260 @default.
• W2972051251 cites W3038058348 @default.
• W2972051251 hasPublicationYear "2019" @default.
• W2972051251 type Work @default.
• W2972051251 sameAs 2972051251 @default.
• W2972051251 citedByCount "24" @default.
• W2972051251 countsByYear W29720512512019 @default.
• W2972051251 countsByYear W29720512512020 @default.
• W2972051251 countsByYear W29720512512021 @default.
• W2972051251 crossrefType "posted-content" @default.
• W2972051251 hasAuthorship W2972051251A5020085226 @default.
• W2972051251 hasAuthorship W2972051251A5039205261 @default.
• W2972051251 hasAuthorship W2972051251A5067919401 @default.
• W2972051251 hasConcept C104317684 @default.
• W2972051251 hasConcept C107673813 @default.
• W2972051251 hasConcept C119857082 @default.
• W2972051251 hasConcept C120665830 @default.
• W2972051251 hasConcept C121332964 @default.
• W2972051251 hasConcept C134306372 @default.
• W2972051251 hasConcept C134752490 @default.
• W2972051251 hasConcept C151730666 @default.
• W2972051251 hasConcept C154945302 @default.
• W2972051251 hasConcept C160920958 @default.
• W2972051251 hasConcept C177769412 @default.
• W2972051251 hasConcept C185592680 @default.
• W2972051251 hasConcept C192209626 @default.
• W2972051251 hasConcept C204321447 @default.
• W2972051251 hasConcept C2779343474 @default.
• W2972051251 hasConcept C33923547 @default.
• W2972051251 hasConcept C36503486 @default.
• W2972051251 hasConcept C37736160 @default.
• W2972051251 hasConcept C41008148 @default.
• W2972051251 hasConcept C44291984 @default.
• W2972051251 hasConcept C55493867 @default.
• W2972051251 hasConcept C63479239 @default.
• W2972051251 hasConcept C86803240 @default.
• W2972051251 hasConceptScore W2972051251C104317684 @default.

• next