GPT-4V (image-caption match answer yes/no, zero-shot)

Reported on 3 benchmarks across 1 task

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