Thanks for this great work! Seems like in the paper, stage 2 is mentioned to have a max context length of 24k which I assume to be a mistake given there are no scripts for that <img width="745" height="152" alt="Image" src="https://github.com/user-attachments/assets/d74bac38-2892-4119-bc9f-ea4915f5c0bb" />
Thanks for this great work!
Seems like in the paper, stage 2 is mentioned to have a max context length of 24k which I assume to be a mistake given there are no scripts for that
