I know this is probably not the case here but I think we should create a standard or way to divide or know based on parameters how many of those affect other la

I know this is probably not the case here but I think we should create a standard or way to divide or know based on parameters how many of those affect other languages. What if it was trained in simplified Chinese and the training/parameters were not sufficient to have it correctly map out that logic to English, so it fails it, yet it can do it correctly in Chinese?