Announcement_4

We released a new sparse attention model for long context, see preprint.