The word theorem comes from late Latin theōrēma and the Greek θεώρημα . If one wanted a plural form other than theorems that reflected its etymology, what would it be? I understand the standard plural is theorems but I would still be interested to know.
As noted in the other answers, "theorems" is the standard usage in math papers. In an odd contrast, however, "lemmata" still has some currency as the plural of "lemma" (an intermediate theorem), and crops up pretty regularly even in very recent papers. On the other hand, my (unresearched) impression is that this too has become less widespread in recent decades.
For comparison, a search for "lemmata" in the text of MathSciNet reviews gives 313 matches, while "theoremata" appears only six times, all in references to the Latin titles of very old papers.
The accepted plural seems to be theorems. At least big dictionaries such as Collins or American Heritage Dic. give no other plural form. For some Latin or Greek word Latin and Greek plural-endings can be found but obviously not for all Latin/Greek words. That would be an over-academization of the vocabulary. And I think it is a good thing that foreign plural-endings are restricted to a limited number of words.
For Latin words the Latin plural is sometimes used, possibly because that language has been the lingua franca in science until very recently, which ensures we are familiar, at least in science, with both the singular and plural form in it's Latin original.
In case of Greek or other origins, I guess the plural reflecting the origins of the word would only be used it that plural denotes an individual concept in English, separate from the singular form.
Two examples I can think of are the Greek
stigmata which is a plural meaning "signs" and the Arabic
Tuareg (a plural meaning "paths" or "ways").