OASIS is a multilingual multimodal QA dataset with 0.92M images, 14.8M QA pairs, and 383 hours of human speech plus cloned audio, built via a semi-automatic framework with human validation to test cultural and commonsense reasoning beyond object recognition.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
OASIS: A Multilingual and Multimodal Dataset for Culturally Grounded Spoken Visual QA
OASIS is a multilingual multimodal QA dataset with 0.92M images, 14.8M QA pairs, and 383 hours of human speech plus cloned audio, built via a semi-automatic framework with human validation to test cultural and commonsense reasoning beyond object recognition.