If captions are meant to be translations, perhaps the example image should show that. Now it looks like it is not really translations at all, but any text string in that language.
Topic on Help talk:File captions
The software is flexible. A caption can be a direct translation, but it can also be a different text string in another language if someone wishes to re-caption the image in another way. It's for the Commons community to decide if it wants a policy around a "standard" caption for translating, or if other captions are permissible.
Of course. I was thinking on the text on this page "Translations can be done.." and "Translations can be added.." but that is not what is shown in the images. That is a bit confusing on a help page for multilingual persons who can tell that the example shows something different than the describing text.
Hi, I would not call it a "translation" unless you really mean it to be one. I think that the text should be the free choice of the person who adds it. Different languages often have speakers with different interests or pre-existing knowledge. Also, I personally find it less collaborative and very restricting if I am supposed to translate 1:1 instead of judging myself what is a proper text in a given language.