Matches in SemOpenAlex for { <https://semopenalex.org/work/W3015468748> ?p ?o ?g. }
- W3015468748 abstract "Transformer-based models are unable to process long sequences due to their self-attention operation, which scales quadratically with the sequence length. To address this limitation, we introduce the Longformer with an attention mechanism that scales linearly with sequence length, making it easy to process documents of thousands of tokens or longer. Longformer's attention mechanism is a drop-in replacement for the standard self-attention and combines a local windowed attention with a task motivated global attention. Following prior work on long-sequence transformers, we evaluate Longformer on character-level language modeling and achieve state-of-the-art results on text8 and enwik8. In contrast to most prior work, we also pretrain Longformer and finetune it on a variety of downstream tasks. Our pretrained Longformer consistently outperforms RoBERTa on long document tasks and sets new state-of-the-art results on WikiHop and TriviaQA. We finally introduce the Longformer-Encoder-Decoder (LED), a Longformer variant for supporting long document generative sequence-to-sequence tasks, and demonstrate its effectiveness on the arXiv summarization dataset." @default.
- W3015468748 created "2020-04-17" @default.
- W3015468748 creator A5016370551 @default.
- W3015468748 creator A5064858748 @default.
- W3015468748 creator A5090038537 @default.
- W3015468748 date "2020-04-10" @default.
- W3015468748 modified "2023-10-06" @default.
- W3015468748 title "Longformer: The Long-Document Transformer" @default.
- W3015468748 cites W1566289585 @default.
- W3015468748 cites W2113459411 @default.
- W3015468748 cites W2130942839 @default.
- W3015468748 cites W2155069789 @default.
- W3015468748 cites W2170973209 @default.
- W3015468748 cites W2338908902 @default.
- W3015468748 cites W2519091744 @default.
- W3015468748 cites W2804032941 @default.
- W3015468748 cites W2805206884 @default.
- W3015468748 cites W2889787757 @default.
- W3015468748 cites W2940744433 @default.
- W3015468748 cites W2946567085 @default.
- W3015468748 cites W2952509486 @default.
- W3015468748 cites W2952809536 @default.
- W3015468748 cites W2962369866 @default.
- W3015468748 cites W2962718483 @default.
- W3015468748 cites W2962739339 @default.
- W3015468748 cites W2963026768 @default.
- W3015468748 cites W2963087868 @default.
- W3015468748 cites W2963088785 @default.
- W3015468748 cites W2963339397 @default.
- W3015468748 cites W2963341956 @default.
- W3015468748 cites W2963403868 @default.
- W3015468748 cites W2963866616 @default.
- W3015468748 cites W2963926728 @default.
- W3015468748 cites W2964110616 @default.
- W3015468748 cites W2965373594 @default.
- W3015468748 cites W2969605360 @default.
- W3015468748 cites W2970120757 @default.
- W3015468748 cites W2970550868 @default.
- W3015468748 cites W2971008823 @default.
- W3015468748 cites W2972324944 @default.
- W3015468748 cites W2972738865 @default.
- W3015468748 cites W2979196189 @default.
- W3015468748 cites W2982399380 @default.
- W3015468748 cites W2984864519 @default.
- W3015468748 cites W2985220278 @default.
- W3015468748 cites W2988421999 @default.
- W3015468748 cites W2994673210 @default.
- W3015468748 cites W2995575179 @default.
- W3015468748 cites W3014438226 @default.
- W3015468748 cites W3015854960 @default.
- W3015468748 cites W3016915903 @default.
- W3015468748 cites W3033182847 @default.
- W3015468748 cites W3034715004 @default.
- W3015468748 cites W3034772996 @default.
- W3015468748 cites W3045733172 @default.
- W3015468748 cites W3082274269 @default.
- W3015468748 cites W3098136301 @default.
- W3015468748 cites W3099876468 @default.
- W3015468748 cites W3105055324 @default.
- W3015468748 cites W3105238007 @default.
- W3015468748 cites W3106298483 @default.
- W3015468748 cites W3131922516 @default.
- W3015468748 hasPublicationYear "2020" @default.
- W3015468748 type Work @default.
- W3015468748 sameAs 3015468748 @default.
- W3015468748 citedByCount "564" @default.
- W3015468748 countsByYear W30154687482019 @default.
- W3015468748 countsByYear W30154687482020 @default.
- W3015468748 countsByYear W30154687482021 @default.
- W3015468748 countsByYear W30154687482022 @default.
- W3015468748 countsByYear W30154687482023 @default.
- W3015468748 crossrefType "posted-content" @default.
- W3015468748 hasAuthorship W3015468748A5016370551 @default.
- W3015468748 hasAuthorship W3015468748A5064858748 @default.
- W3015468748 hasAuthorship W3015468748A5090038537 @default.
- W3015468748 hasConcept C111919701 @default.
- W3015468748 hasConcept C118505674 @default.
- W3015468748 hasConcept C119599485 @default.
- W3015468748 hasConcept C127413603 @default.
- W3015468748 hasConcept C154945302 @default.
- W3015468748 hasConcept C165801399 @default.
- W3015468748 hasConcept C170858558 @default.
- W3015468748 hasConcept C204321447 @default.
- W3015468748 hasConcept C2778112365 @default.
- W3015468748 hasConcept C28490314 @default.
- W3015468748 hasConcept C39890363 @default.
- W3015468748 hasConcept C41008148 @default.
- W3015468748 hasConcept C54355233 @default.
- W3015468748 hasConcept C66322947 @default.
- W3015468748 hasConcept C86803240 @default.
- W3015468748 hasConceptScore W3015468748C111919701 @default.
- W3015468748 hasConceptScore W3015468748C118505674 @default.
- W3015468748 hasConceptScore W3015468748C119599485 @default.
- W3015468748 hasConceptScore W3015468748C127413603 @default.
- W3015468748 hasConceptScore W3015468748C154945302 @default.
- W3015468748 hasConceptScore W3015468748C165801399 @default.
- W3015468748 hasConceptScore W3015468748C170858558 @default.
- W3015468748 hasConceptScore W3015468748C204321447 @default.
- W3015468748 hasConceptScore W3015468748C2778112365 @default.
- W3015468748 hasConceptScore W3015468748C28490314 @default.