Prompting Large Vision-Language Models for Compositional Reasoning

Open paper
Benchmark Model Rank Results
visual-reasoning-on-winogroundKeyComp* (GPT-4)#21Text Score: 43.5Image Score: 28.7Group Score: 18.2
visual-reasoning-on-winogroundKeyComp* (GPT-3.5)#22Text Score: 42.7Image Score: 27.8Group Score: 17.4
visual-reasoning-on-winogroundKeyComp (GPT-3.5)#46Text Score: 30.3Image Score: 24.6Group Score: 12.4

Rank counts only results with a code link.