Some questions about your paper

[zinuoli]

Hi, thank you very much for this great and valuable work!

I got some questions when reading your paper, it mentioned VALOR-32K are splited into 25K/3.5K/3.5K videos for training, validation and testing, respectively., but in your paper it seems you didn't use the 25K samples in VALOR-32K.

May I know how do you benchmark your model on VALOR-32K? Solely on 3.5K for testing or you use VALOR-32K for training and validation as well?

Thank you again : )