Breaking Common Sense: WHOOPS! A Vision-and-Language Benchmark of Synthetic and Compositional Images

Keyphrases

• Vision-language 100%
• AI Models 100%
• Image Generation 50%
• Publicly Available 50%
• New Dataset 50%
• Human Performance 50%
• State-of-the-art Models 50%
• Lag behind 50%
• Reasoning Ability 50%
• New Benchmark 50%
• Cross-modal Matching 50%
• Soccer 50%
• Chess 50%
• GPT-3 50%
• Generation Task 50%
• Visual Question Answering 50%
• Midjourney 50%
• Football 50%
• Lionel Messi 50%
• 2022 World Cup 50%
• Image Captioning 50%
• Explanation Generation 50%
• Cristiano Ronaldo 50%
• Generation Tools 50%
• Visual Commonsense Reasoning 50%

Computer Science

• Common Sense 100%
• Artificial Intelligence 100%
• Image Synthesis 50%
• Human Performance 50%
• Commonsense Reasoning 50%
• Visual Question Answering 50%
• Image Captioning 50%