Questions about experimental results

[chaochen99]

Hi，

The experimental results on the LSDMC dataset in the paper reach a very high level, even higher than the models initialized with CLIP, such as x-pool, CenterCLIP, etc. However, the MSRVTT dataset results are generally several points lower than these models. I wonder if this is a reasonable phenomenon?

Looking forward to your reply.

Thanks!