Questions about Training Data

[RuixiangZhao]

Hi authors, thanks for the interesting work. I have two quick questions about the training data:

• Do you plan to release the training datasets (or part of them)?

• Section 4.1 mainly lists the data sources. Could you clarify how the training data are constructed, e.g., how the 50K implicit QA pairs and 122K streaming video QA pairs are curated?

Thanks for your time!