>>105687416
>did they release the math shit on the paper? if yes maybe we can use it to code our own sage2++?
I'm not sure if they showed all the details but yes, they talked abut the specifics on how they achieved this speed improvement
https://arxiv.org/pdf/2505.21136