Yeah — simply multiplied rope * 2, trained on long data with a MSL of 16K.

→ View original post on X — @mattshumer_