>>106554001
>Do we have any guarantee that we aren't just running the model wrong?
it doesn't seem to ask for something specific, I hope I'm wrong and in reality the model can do better than that
https://github.com/Tencent-Hunyuan/SRPO?tab=readme-ov-file#inference