Maybe it's the exposure to Chinese? I've heard that training models on code first helps, so I could see it. I've also heard hearsay that R1 is quite clever in

Maybe it's the exposure to Chinese? I've heard that training models on code first helps, so I could see it. I've also heard hearsay that R1 is quite clever in Chinese, too.