chinese is less token efficient than english
How so? Do you mean in practice when using mostly english trained models?
alt Hacker News
How so? Do you mean in practice when using mostly english trained models?