The unspoken assumption is that if the model reasons well in English, surely it reasons almost as well everywhere else. It does not.
The unspoken assumption is that if the model reasons well in English, surely it reasons almost as well everywhere else. It does not.