single mention · not yet filed
Tri-hybrid layer allocation with 13 convolution, 5 sliding-window attention, and 6 global-attention layers
The sources on this page only mention it in passing.
single mention · not yet filed
The sources on this page only mention it in passing.