Visual Question Answering (VQA) on WHOOPS!

Metric: Exact Match (higher is better)

LeaderboardDataset
Loading chart...
#ModelExact MatchExtra DataPaperDateCode
1BLIP2 FlanT5-XXL (Fine-tuned)21YesBreaking Common Sense: WHOOPS! A Vision-and-Lang...2023-03-13-
2BLIP2 FlanT5-XL (Fine-tuned)20YesBreaking Common Sense: WHOOPS! A Vision-and-Lang...2023-03-13-
3BLIP2 FlanT5-XXL (Zero-shot)15NoBreaking Common Sense: WHOOPS! A Vision-and-Lang...2023-03-13-
4OFA Large8NoBreaking Common Sense: WHOOPS! A Vision-and-Lang...2023-03-13-
5BLIP Large6NoBreaking Common Sense: WHOOPS! A Vision-and-Lang...2023-03-13-
6BLIP2 FlanT5-XXL (Text-only FT)4YesBreaking Common Sense: WHOOPS! A Vision-and-Lang...2023-03-13-