Can SAM3 use both initial point information and prompt semantic information during the inference process?

Hello, thank you very much for your work. Can SAM3 use both initial point information and prompt semantic information during the inference process?
I encountered the following issues while tracking objects of the same type:
1. Multiple objects of the same category can be completely segmented using only prompt semantic information, but the IDs between objects may change
2. Therefore, I attempted to add initial segmentation points for these objects separately, while also incorporating prompt semantic information. However, I found that some of the objects could not be segmented, and the effect was much worse than a simple prompt.
Thank you very much for your reply!