Matches in SemOpenAlex for { <https://semopenalex.org/work/W3150658752> ?p ?o ?g. }
Showing items 1 to 89 of
89
with 100 items per page.
- W3150658752 abstract "We introduce an evaluation methodology for visual question answering (VQA) to better diagnose cases of shortcut learning. These cases happen when a model exploits spurious statistical regularities to produce correct answers but does not actually deploy the desired behavior. There is a need to identify possible shortcuts in a dataset and assess their use before deploying a model in the real world. The research community in VQA has focused exclusively on question-based shortcuts, where a model might, for example, answer What is the color of the sky with blue by relying mostly on the question-conditional training prior and give little weight to visual evidence. We go a step further and consider multimodal shortcuts that involve both questions and images. We first identify potential shortcuts in the popular VQA v2 training set by mining trivial predictive rules such as co-occurrences of words and visual elements. We then introduce VQA-CounterExamples (VQACE), an evaluation protocol based on our subset of CounterExamples i.e. image-question-answer triplets where our rules lead to incorrect answers. We use this new evaluation in a large-scale study of existing approaches for VQA. We demonstrate that even state-of-the-art models perform poorly and that existing techniques to reduce biases are largely ineffective in this context. Our findings suggest that past work on question-based biases in VQA has only addressed one facet of a complex issue. The code for our method is available at https://github.com/cdancette/detect-shortcuts" @default.
- W3150658752 created "2021-04-13" @default.
- W3150658752 creator A5022871131 @default.
- W3150658752 creator A5036295862 @default.
- W3150658752 creator A5046767053 @default.
- W3150658752 creator A5067549788 @default.
- W3150658752 date "2021-11-11" @default.
- W3150658752 modified "2023-09-27" @default.
- W3150658752 title "Beyond Question-Based Biases: Assessing Multimodal Shortcut Learning in Visual Question Answering" @default.
- W3150658752 cites W1861492603 @default.
- W3150658752 cites W1933349210 @default.
- W3150658752 cites W2171810632 @default.
- W3150658752 cites W2282821441 @default.
- W3150658752 cites W2560730294 @default.
- W3150658752 cites W2561715562 @default.
- W3150658752 cites W2597425697 @default.
- W3150658752 cites W2608030593 @default.
- W3150658752 cites W2745461083 @default.
- W3150658752 cites W2789240707 @default.
- W3150658752 cites W2886641317 @default.
- W3150658752 cites W2895472239 @default.
- W3150658752 cites W2902617128 @default.
- W3150658752 cites W2940753425 @default.
- W3150658752 cites W2951286828 @default.
- W3150658752 cites W2962685807 @default.
- W3150658752 cites W2962884579 @default.
- W3150658752 cites W2963224792 @default.
- W3150658752 cites W2963518342 @default.
- W3150658752 cites W2963644680 @default.
- W3150658752 cites W2963717374 @default.
- W3150658752 cites W2963969878 @default.
- W3150658752 cites W2964072591 @default.
- W3150658752 cites W2964118342 @default.
- W3150658752 cites W2965628639 @default.
- W3150658752 cites W2968993450 @default.
- W3150658752 cites W2970017794 @default.
- W3150658752 cites W2970019270 @default.
- W3150658752 cites W2970115835 @default.
- W3150658752 cites W2970608575 @default.
- W3150658752 cites W2970692043 @default.
- W3150658752 cites W2982699810 @default.
- W3150658752 cites W3007641164 @default.
- W3150658752 cites W3016970897 @default.
- W3150658752 cites W3026376263 @default.
- W3150658752 cites W3034564653 @default.
- W3150658752 cites W3035517717 @default.
- W3150658752 cites W3035561630 @default.
- W3150658752 cites W3098528040 @default.
- W3150658752 cites W3100511085 @default.
- W3150658752 cites W3101609372 @default.
- W3150658752 cites W3104788521 @default.
- W3150658752 cites W3104219743 @default.
- W3150658752 hasPublicationYear "2021" @default.
- W3150658752 type Work @default.
- W3150658752 sameAs 3150658752 @default.
- W3150658752 citedByCount "4" @default.
- W3150658752 countsByYear W31506587522021 @default.
- W3150658752 crossrefType "proceedings-article" @default.
- W3150658752 hasAuthorship W3150658752A5022871131 @default.
- W3150658752 hasAuthorship W3150658752A5036295862 @default.
- W3150658752 hasAuthorship W3150658752A5046767053 @default.
- W3150658752 hasAuthorship W3150658752A5067549788 @default.
- W3150658752 hasBestOaLocation W31506587522 @default.
- W3150658752 hasConcept C107457646 @default.
- W3150658752 hasConcept C154945302 @default.
- W3150658752 hasConcept C41008148 @default.
- W3150658752 hasConcept C44291984 @default.
- W3150658752 hasConceptScore W3150658752C107457646 @default.
- W3150658752 hasConceptScore W3150658752C154945302 @default.
- W3150658752 hasConceptScore W3150658752C41008148 @default.
- W3150658752 hasConceptScore W3150658752C44291984 @default.
- W3150658752 hasLocation W31506587521 @default.
- W3150658752 hasLocation W31506587522 @default.
- W3150658752 hasOpenAccess W3150658752 @default.
- W3150658752 hasPrimaryLocation W31506587521 @default.
- W3150658752 hasRelatedWork W105002793 @default.
- W3150658752 hasRelatedWork W128392744 @default.
- W3150658752 hasRelatedWork W1550833313 @default.
- W3150658752 hasRelatedWork W1940793384 @default.
- W3150658752 hasRelatedWork W2204505259 @default.
- W3150658752 hasRelatedWork W2341207148 @default.
- W3150658752 hasRelatedWork W2351286801 @default.
- W3150658752 hasRelatedWork W2354866896 @default.
- W3150658752 hasRelatedWork W2361152157 @default.
- W3150658752 hasRelatedWork W2805599431 @default.
- W3150658752 isParatext "false" @default.
- W3150658752 isRetracted "false" @default.
- W3150658752 magId "3150658752" @default.
- W3150658752 workType "article" @default.