Hi, thanks for your work.
I would like to ask about what does the field "time" of breakpoint mode omean in your json file?
And when you evaluate LLM-based video understanding models like VideoChat, does breakpoint mode act more like single-image perception rather than long-video understanding?
thanks!
Hi, thanks for your work.
I would like to ask about what does the field "time" of breakpoint mode omean in your json file?
And when you evaluate LLM-based video understanding models like VideoChat, does breakpoint mode act more like single-image perception rather than long-video understanding?
thanks!