Matches in SemOpenAlex for { <https://semopenalex.org/work/W3204669968> ?p ?o ?g. }
- W3204669968 abstract "Many datasets have been created for training reading comprehension models, and a natural question is whether we can combine them to build models that (1) perform better on all of the training datasets and (2) generalize and transfer better to new datasets. Prior work has addressed this goal by training one network simultaneously on multiple datasets, which works well on average but is prone to over- or under-fitting different sub- distributions and might transfer worse compared to source models with more overlap with the target dataset. Our approach is to model multi-dataset question answering with an ensemble of single-dataset experts, by training a collection of lightweight, dataset-specific adapter modules (Houlsby et al., 2019) that share an underlying Transformer model. We find that these Multi-Adapter Dataset Experts (MADE) outperform all our baselines in terms of in-distribution accuracy, and simple methods based on parameter-averaging lead to better zero-shot generalization and few-shot transfer performance, offering a strong and versatile starting point for building new reading comprehension systems." @default.
- W3204669968 created "2021-10-11" @default.
- W3204669968 creator A5005009703 @default.
- W3204669968 creator A5051064208 @default.
- W3204669968 creator A5071180534 @default.
- W3204669968 date "2021-01-01" @default.
- W3204669968 modified "2023-09-27" @default.
- W3204669968 title "Single-dataset Experts for Multi-dataset Question Answering" @default.
- W3204669968 cites W1981208470 @default.
- W3204669968 cites W2427527485 @default.
- W3204669968 cites W2481265265 @default.
- W3204669968 cites W2557764419 @default.
- W3204669968 cites W2606964149 @default.
- W3204669968 cites W2609826708 @default.
- W3204669968 cites W2746097825 @default.
- W3204669968 cites W2799054028 @default.
- W3204669968 cites W2889787757 @default.
- W3204669968 cites W2908510526 @default.
- W3204669968 cites W2911300548 @default.
- W3204669968 cites W2912924812 @default.
- W3204669968 cites W2919420119 @default.
- W3204669968 cites W2947567669 @default.
- W3204669968 cites W2951036431 @default.
- W3204669968 cites W2962881743 @default.
- W3204669968 cites W2963339397 @default.
- W3204669968 cites W2963341956 @default.
- W3204669968 cites W2963928014 @default.
- W3204669968 cites W2964303773 @default.
- W3204669968 cites W2965373594 @default.
- W3204669968 cites W2979826702 @default.
- W3204669968 cites W2986532682 @default.
- W3204669968 cites W2988421999 @default.
- W3204669968 cites W3034407524 @default.
- W3204669968 cites W3099655892 @default.
- W3204669968 cites W3099793224 @default.
- W3204669968 cites W3118850317 @default.
- W3204669968 cites W3153675281 @default.
- W3204669968 doi "https://doi.org/10.18653/v1/2021.emnlp-main.495" @default.
- W3204669968 hasPublicationYear "2021" @default.
- W3204669968 type Work @default.
- W3204669968 sameAs 3204669968 @default.
- W3204669968 citedByCount "3" @default.
- W3204669968 countsByYear W32046699682022 @default.
- W3204669968 countsByYear W32046699682023 @default.
- W3204669968 crossrefType "proceedings-article" @default.
- W3204669968 hasAuthorship W3204669968A5005009703 @default.
- W3204669968 hasAuthorship W3204669968A5051064208 @default.
- W3204669968 hasAuthorship W3204669968A5071180534 @default.
- W3204669968 hasBestOaLocation W32046699681 @default.
- W3204669968 hasConcept C119857082 @default.
- W3204669968 hasConcept C121332964 @default.
- W3204669968 hasConcept C124101348 @default.
- W3204669968 hasConcept C134306372 @default.
- W3204669968 hasConcept C150899416 @default.
- W3204669968 hasConcept C154945302 @default.
- W3204669968 hasConcept C165801399 @default.
- W3204669968 hasConcept C177148314 @default.
- W3204669968 hasConcept C17744445 @default.
- W3204669968 hasConcept C199539241 @default.
- W3204669968 hasConcept C2778780117 @default.
- W3204669968 hasConcept C33923547 @default.
- W3204669968 hasConcept C41008148 @default.
- W3204669968 hasConcept C44291984 @default.
- W3204669968 hasConcept C51632099 @default.
- W3204669968 hasConcept C554936623 @default.
- W3204669968 hasConcept C62520636 @default.
- W3204669968 hasConcept C66322947 @default.
- W3204669968 hasConceptScore W3204669968C119857082 @default.
- W3204669968 hasConceptScore W3204669968C121332964 @default.
- W3204669968 hasConceptScore W3204669968C124101348 @default.
- W3204669968 hasConceptScore W3204669968C134306372 @default.
- W3204669968 hasConceptScore W3204669968C150899416 @default.
- W3204669968 hasConceptScore W3204669968C154945302 @default.
- W3204669968 hasConceptScore W3204669968C165801399 @default.
- W3204669968 hasConceptScore W3204669968C177148314 @default.
- W3204669968 hasConceptScore W3204669968C17744445 @default.
- W3204669968 hasConceptScore W3204669968C199539241 @default.
- W3204669968 hasConceptScore W3204669968C2778780117 @default.
- W3204669968 hasConceptScore W3204669968C33923547 @default.
- W3204669968 hasConceptScore W3204669968C41008148 @default.
- W3204669968 hasConceptScore W3204669968C44291984 @default.
- W3204669968 hasConceptScore W3204669968C51632099 @default.
- W3204669968 hasConceptScore W3204669968C554936623 @default.
- W3204669968 hasConceptScore W3204669968C62520636 @default.
- W3204669968 hasConceptScore W3204669968C66322947 @default.
- W3204669968 hasLocation W32046699681 @default.
- W3204669968 hasLocation W32046699682 @default.
- W3204669968 hasOpenAccess W3204669968 @default.
- W3204669968 hasPrimaryLocation W32046699681 @default.
- W3204669968 hasRelatedWork W2946016983 @default.
- W3204669968 hasRelatedWork W2960456850 @default.
- W3204669968 hasRelatedWork W3216313460 @default.
- W3204669968 hasRelatedWork W4213299466 @default.
- W3204669968 hasRelatedWork W4281382123 @default.
- W3204669968 hasRelatedWork W4281645081 @default.
- W3204669968 hasRelatedWork W4294306704 @default.
- W3204669968 hasRelatedWork W4308262314 @default.
- W3204669968 hasRelatedWork W4318834068 @default.
- W3204669968 hasRelatedWork W4318957922 @default.
- W3204669968 isParatext "false" @default.