LaDiC: Are Diffusion Models Really Inferior to Autoregressive Counterparts for Image-to-Text Generation?

Open paper
Benchmark Model Rank Results
image-captioning-on-coco-captionsLaDiC (ours, 30 steps)#31BLEU-4: 0.382CIDER: 126.2METEOR: 29.5
image-captioning-on-coco-captionsLaDiC#35SPICE: 22.4ROUGE-L: 58.7

Rank counts only results with a code link.