BLIP2 FlanT5-XXL (Zero-shot)

Reported on 4 benchmarks across 3 tasks · 1 paper

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Natural Language Processing4 results

Visual Question Answering (VQA)onWHOOPS!
BEM· 2023-03-13
55
best: 57 (BLIP2 FlanT5-XXL (Fine-tuned))
Breaking Common Sense: WHOOPS! A Vision-and-Language Benchmark of Synthetic and Compositional Images arXiv:2303.07274
Visual Question Answering (VQA)onWHOOPS!
Exact Match· 2023-03-13
15
best: 21 (BLIP2 FlanT5-XXL (Fine-tuned))
Breaking Common Sense: WHOOPS! A Vision-and-Language Benchmark of Synthetic and Compositional Images arXiv:2303.07274
Image-to-Text RetrievalonWHOOPS!
Specificity· 2023-03-13
71
best: 94 (BLIP2 FlanT5-XXL (Text-only FT))
Breaking Common Sense: WHOOPS! A Vision-and-Language Benchmark of Synthetic and Compositional Images arXiv:2303.07274
Explanation GenerationonWHOOPS!
Human (%)
0
best: 68 (Ground-truth Caption -> GPT3 (Oracle))