GPT Models Fail to Find Language Structure in the Voynich Manuscript

[obelus]

I trained several nanoGPT models

Is it possible that n-gram predictability is driven by the absolute frequency of the most-frequent words?  In Early Modern English (I checked the 1611 King James Gospels), the most frequent word ("and", "the", "of", "he", "him", ...) occurs ~3 times more often than the first-rank word in Cuva paragraph text ("DAM", "SEDY", "OL", "AM", "ZEDY", ...):

   
The normalized Cuva frequency distribution (blue) is wider than the English one (magenta).  If these frequency values are correlated with n-gram predictability, then (based on your results) we would expect the output of Torsten's generator to have a profile that more closely resembles English.  A sample of 34 000 words, generated with the default parameters and expressed as Cuva, are represented by grey stars ("O", "SUDY", "HO", "Y", "AM", ...).

It makes combinatorial sense that n-grams are easier to predict in a sample where frequency is concentrated in fewer words.  I can not judge whether any further explanation is required, because the workings of the transformer are obscure to me.  Do we know what properties specifically get vectorized?