WinoGAViL
Github
Huggingface
Leaderboard
Paper
Explore
Challenge the AI
Download
WinoGAViL: Gamified Association Benchmark to Challenge Vision-and-Language Models
Five images are presented below.
A cue word and number of associated images are given.
Choose the images that best associates with the cue word.
Cue
Associations
2
AI Response
👉
Choose here!
Go to
Challenge the AI
page to play the associations games against a rival AI
3.5K+
Associations
<52%
Model Success Rate
>90%
Human Successes Rate
Paper
Authors
Yonatan Bitton
Nitzan Guetta
Ron Yosef
Yuval Elovici
Mohit Bansal
Gabriel Stanovsky
Roy Schwartz