Does the top-two captions per image included in BLIP2 training benchmark datasets?

[inkzk]

According to the BLIP2 paper:

We adopt the CapFilt method (Li et al., 2022) to create synthetic captions for the web images... We keep top-two captions per image as training data and randomly sample one at each pre-training step.

Thus I wonder wether the top-two captions included in the training dataset of benchmark link, Further more, the LAION400M dataset not found in benchmark link, any idea?

[LiJunnan1992]

We have not yet released the synthetic captions for LAION. You may consider to generate them yourself if needed. Thanks.

[inkzk]

We have not yet released the synthetic captions for LAION. You may consider to generate them yourself if needed. Thanks.

Thanks for reply. Does the synthetic captions for others datasets released?