MetaLM

Reported on 6 benchmarks across 2 tasks · 1 paper · 2 SOTA

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

Natural Language Processing6 results

Image Captioningonnocaps val
CIDEr· 2022-06-13
58.7
best: 107.9 (Prismer)
SOTA
Language Models are General-Purpose Interfaces arXiv:2206.06336
Image Captioningonnocaps val
SPICE· 2022-06-13
8.6
best: 14.8 (Prismer)
SOTA
Language Models are General-Purpose Interfaces arXiv:2206.06336
Visual Question Answering (VQA)onOK-VQA
Accuracy· 2022-06-13
11.4
best: 66.8 (PaLI-X-VPD)
Language Models are General-Purpose Interfaces arXiv:2206.06336
Visual Question Answering (VQA)onVQA v2 val
Accuracy· 2022-06-13
41.1
best: 82.19 (BLIP-2 ViT-G OPT 6.7B (fine-tuned))
Language Models are General-Purpose Interfaces arXiv:2206.06336
Image CaptioningonFlickr30k Captions test
CIDEr· 2022-06-13
43.3
best: 67.4 (Unified VLP)
Language Models are General-Purpose Interfaces arXiv:2206.06336
Image CaptioningonFlickr30k Captions test
SPICE· 2022-06-13
11.7
best: 17 (Unified VLP)
Language Models are General-Purpose Interfaces arXiv:2206.06336